Prepare release of 0.30.0

Remove Channel::request_cancellation.
This trait fn returns a private type, which means it's useless for anyone using the Channel. Instead, add an inert (now-public) ResponseGuard to TrackedRequest that, when taken out of the ManuallyDrop, ensures a Channel's request state is cleaned up. It's preferable to make ResponseGuard public instead of RequestCancellations because it's a smaller API surface (no public methods, just a Drop fn) and harder to misuse, because it is already associated with the correct request ID to cancel.
2026-02-23 15:49:54 +01:00 · 2022-08-12 16:08:33 -07:00 · 2022-08-12 16:08:33 -07:00 · 2022-08-12 09:04:47 -07:00 · 2022-08-10 17:11:50 -07:00 · 2022-07-15 10:14:49 -07:00
51 changed files with 2894 additions and 1744 deletions
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -1,7 +1,11 @@
 [workspace]
+resolver = "2"

 members = [
    "example-service",
    "tarpc",
    "plugins",
 ]
+
+[profile.dev]
+split-debuginfo = "unpacked"
--- a/README.md
+++ b/README.md
@@ -40,7 +40,7 @@ rather than in a separate language such as .proto. This means there's no separat
 process, and no context switching between different languages.

 Some other features of tarpc:
- Pluggable transport: any type impling `Stream<Item = Request> + Sink<Response>` can be
+- Pluggable transport: any type implementing `Stream<Item = Request> + Sink<Response>` can be
  used as a transport to connect the client and server.
 - `Send + 'static` optional: if the transport doesn't require it, neither does tarpc!
 - Cascading cancellation: dropping a request will send a cancellation message to the server.
@@ -51,6 +51,14 @@ Some other features of tarpc:
  requests sent by the server that use the request context will propagate the request deadline.
  For example, if a server is handling a request with a 10s deadline, does 2s of work, then
  sends a request to another server, that server will see an 8s deadline.
+- Distributed tracing: tarpc is instrumented with
+  [tracing](https://github.com/tokio-rs/tracing) primitives extended with
+  [OpenTelemetry](https://opentelemetry.io/) traces. Using a compatible tracing subscriber like
+  [Jaeger](https://github.com/open-telemetry/opentelemetry-rust/tree/main/opentelemetry-jaeger),
+  each RPC can be traced through the client, server, and other dependencies downstream of the
+  server. Even for applications not connected to a distributed tracing collector, the
+  instrumentation can also be ingested by regular loggers like
+  [env_logger](https://github.com/env-logger-rs/env_logger/).
 - Serde serialization: enabling the `serde1` Cargo feature will make service requests and
  responses `Serialize + Deserialize`. It's entirely optional, though: in-memory transports can
  be used, as well, so the price of serialization doesn't have to be paid when it's not needed.
@@ -59,7 +67,7 @@ Some other features of tarpc:
 Add to your `Cargo.toml` dependencies:

 ```toml
-tarpc = "0.25"
+tarpc = "0.29"
 ```

 The `tarpc::service` attribute expands to a collection of items that form an rpc service.
@@ -72,8 +80,9 @@ This example uses [tokio](https://tokio.rs), so add the following dependencies t
 your `Cargo.toml`:

 ```toml
-futures = "1.0"
-tarpc = { version = "0.25", features = ["tokio1"] }
+anyhow = "1.0"
+futures = "0.3"
+tarpc = { version = "0.29", features = ["tokio1"] }
 tokio = { version = "1.0", features = ["macros"] }
 ```

@@ -91,9 +100,8 @@ use futures::{
 };
 use tarpc::{
    client, context,
-    server::{self, Incoming},
+    server::{self, incoming::Incoming, Channel},
 };
-use std::io;

 // This is the service definition. It looks a lot like a trait definition.
 // It defines one RPC, hello, which takes one arg, name, and returns a String.
@@ -120,7 +128,7 @@ impl World for HelloServer {
    type HelloFut = Ready<String>;

    fn hello(self, _: context::Context, name: String) -> Self::HelloFut {
-        future::ready(format!("Hello, {}!", name))
+        future::ready(format!("Hello, {name}!"))
    }
 }
 ```
@@ -132,7 +140,7 @@ available behind the `tcp` feature.

 ```rust
 #[tokio::main]
-async fn main() -> io::Result<()> {
+async fn main() -> anyhow::Result<()> {
    let (client_transport, server_transport) = tarpc::transport::channel::unbounded();

    let server = server::BaseChannel::with_defaults(server_transport);
@@ -140,14 +148,14 @@ async fn main() -> io::Result<()> {

    // WorldClient is generated by the #[tarpc::service] attribute. It has a constructor `new`
    // that takes a config and any Transport as input.
-    let mut client = WorldClient::new(client::Config::default(), client_transport).spawn()?;
+    let mut client = WorldClient::new(client::Config::default(), client_transport).spawn();

    // The client has an RPC method for each RPC defined in the annotated trait. It takes the same
    // args as defined, with the addition of a Context, which is always the first arg. The Context
    // specifies a deadline and trace information which can be helpful in debugging requests.
    let hello = client.hello(context::current(), "Stim".to_string()).await?;

-    println!("{}", hello);
+    println!("{hello}");

    Ok(())
 }
--- a/RELEASES.md
+++ b/RELEASES.md
@@ -1,3 +1,152 @@
+## 0.30.0 (2022-08-12)
+
+### Breaking Changes
+
+- Some types that impl Future are now annotated with `#[must_use]`. Code that previously created
+  these types but did not use them will now receive a warning. Code that disallows warnings will
+  receive a compilation error.
+
+### Fixes
+
+- Servers will more reliably clean up request state for requests with long deadlines when response
+processing is aborted without sending a response.
+
+### Other Changes
+
+- `TrackedRequest` now contains a response guard that can be used to ensure state cleanup for
+  aborted requests. (This was already handled automatically by `InFlightRequests`).
+- When the feature serde-transport is enabled, the crate tokio_serde is now re-exported.
+
+## 0.29.0 (2022-05-26)
+
+### Breaking Changes
+
+`Context.deadline` is now serialized as a Duration. This prevents clock skew from affecting deadline
+behavior. For more details see https://github.com/google/tarpc/pull/367 and its [related
+issue](https://github.com/google/tarpc/issues/366).
+
+## 0.28.0 (2022-04-06)
+
+### Breaking Changes
+
+- The minimum supported Rust version has increased to 1.58.0.
+- The version of opentelemetry depended on by tarpc has increased to 0.17.0.
+
+## 0.27.2 (2021-10-08)
+
+### Fixes
+
+Clients will now close their transport before dropping it. An attempt at a clean shutdown can help
+the server drop its connections more quickly.
+
+## 0.27.1 (2021-09-22)
+
+### Breaking Changes
+
+#### RPC error type is changing
+
+RPC return types are changing from `Result<Response, io::Error>` to `Result<Response,
+tarpc::client::RpcError>`.
+
+Becaue tarpc is a library, not an application, it should strive to
+use structured errors in its API so that users have maximal flexibility
+in how they handle errors. io::Error makes that hard, because it is a
+kitchen-sink error type.
+
+RPCs in particular only have 3 classes of errors:
+
+- The connection breaks.
+- The request expires.
+- The server decides not to process the request.
+
+RPC responses can also contain application-specific errors, but from the
+perspective of the RPC library, those are opaque to the framework, classified
+as successful responsees.
+
+### Open Telemetry
+
+The Opentelemetry dependency is updated to version 0.16.x.
+
+## 0.27.0 (2021-09-22)
+
+This version was yanked due to tarpc-plugins version mismatches.
+
+
+## 0.26.0 (2021-04-14)
+
+### New Features
+
+#### Tracing
+
+tarpc is now instrumented with tracing primitives extended with
+OpenTelemetry traces. Using a compatible tracing-opentelemetry
+subscriber like Jaeger, each RPC can be traced through the client,
+server, amd other dependencies downstream of the server. Even for
+applications not connected to a distributed tracing collector, the
+instrumentation can also be ingested by regular loggers like env_logger.
+
+### Breaking Changes
+
+#### Logging
+
+Logged events are now structured using tracing. For applications using a
+logger and not a tracing subscriber, these logs may look different or
+contain information in a less consumable manner. The easiest solution is
+to add a tracing subscriber that logs to stdout, such as
+tracing_subscriber::fmt.
+
+####  Context
+
+- Context no longer has parent_span, which was actually never needed,
+  because the context sent in an RPC is inherently the parent context.
+  For purposes of distributed tracing, the client side of the RPC has all
+  necessary information to link the span to its parent; the server side
+  need do nothing more than export the (trace ID, span ID) tuple.
+- Context has a new field, SamplingDecision, which has two variants,
+  Sampled and Unsampled. This field can be used by downstream systems to
+  determine whether a trace needs to be exported. If the parent span is
+  sampled, the expectation is that all child spans be exported, as well;
+  to do otherwise could result in lossy traces being exported. Note that
+  if an Openetelemetry tracing subscriber is not installed, the fallback
+  context will still be used, but the Context's sampling decision will
+  always be inherited by the parent Context's sampling decision.
+- Context::scope has been removed. Context propagation is now done via
+  tracing's task-local spans. Spans can be propagated across tasks via
+  Span::in_scope. When a service receives a request, it attaches an
+  Opentelemetry context to the local Span created before request handling,
+  and this context contains the request deadline. This span-local deadline
+  is retrieved by Context::current, but it cannot be modified so that
+  future Context::current calls contain a different deadline. However, the
+  deadline in the context passed into an RPC call will override it, so
+  users can retrieve the current context and then modify the deadline
+  field, as has been historically possible.
+- Context propgation precedence changes: when an RPC is initiated, the
+  current Span's Opentelemetry context takes precedence over the trace
+  context passed into the RPC method. If there is no current Span, then
+  the trace context argument is used as it has been historically. Note
+  that Opentelemetry context propagation requires an Opentelemetry
+  tracing subscriber to be installed.
+
+#### Server
+
+- The server::Channel trait now has an additional required associated
+  type and method which returns the underlying transport. This makes it
+  more ergonomic for users to retrieve transport-specific information,
+  like IP Address. BaseChannel implements Channel::transport by returning
+  the underlying transport, and channel decorators like Throttler just
+  delegate to the Channel::transport method of the wrapped channel.
+
+#### Client
+
+- NewClient::spawn no longer returns a result, as spawn can't fail.
+
+### References
+
+1. https://github.com/tokio-rs/tracing
+2. https://opentelemetry.io
+3. https://github.com/open-telemetry/opentelemetry-rust/tree/main/opentelemetry-jaeger
+4. https://github.com/env-logger-rs/env_logger
+
 ## 0.25.0 (2021-03-10)

 ### Breaking Changes
@@ -170,7 +319,7 @@ nameable futures and will just be boxing the return type anyway. This macro does

 ### Breaking Changes

- Enums had _non_exhaustive fields replaced with the #[non_exhaustive] attribute.
+- Enums had `_non_exhaustive` fields replaced with the #[non_exhaustive] attribute.

 ### Bug Fixes

--- a/example-service/Cargo.toml
+++ b/example-service/Cargo.toml
@@ -1,8 +1,9 @@
 [package]
 name = "tarpc-example-service"
-version = "0.9.0"
+version = "0.12.0"
+rust-version = "1.56"
 authors = ["Tim Kuehn <tikue@google.com>"]
-edition = "2018"
+edition = "2021"
 license = "MIT"
 documentation = "https://docs.rs/tarpc-example-service"
 homepage = "https://github.com/google/tarpc"
@@ -13,12 +14,18 @@ readme = "../README.md"
 description = "An example server built on tarpc."

 [dependencies]
-clap = "2.33"
-env_logger = "0.8"
+anyhow = "1.0"
+clap = { version = "3.0.0-rc.9", features = ["derive"] }
+log = "0.4"
 futures = "0.3"
-serde = { version = "1.0" }
-tarpc = { version = "0.25", path = "../tarpc", features = ["full"] }
-tokio = { version = "1", features = ["full"] }
+opentelemetry = { version = "0.17", features = ["rt-tokio"] }
+opentelemetry-jaeger = { version = "0.16", features = ["rt-tokio"] }
+rand = "0.8"
+tarpc = { version = "0.30", path = "../tarpc", features = ["full"] }
+tokio = { version = "1", features = ["macros", "net", "rt-multi-thread"] }
+tracing = { version = "0.1" }
+tracing-opentelemetry = "0.17"
+tracing-subscriber = {version = "0.3", features = ["env-filter"]}

 [lib]
 name = "service"
--- a/example-service/src/client.rs
+++ b/example-service/src/client.rs
@@ -4,57 +4,49 @@
 // license that can be found in the LICENSE file or at
 // https://opensource.org/licenses/MIT.

-use clap::{App, Arg};
-use std::{io, net::SocketAddr};
+use clap::Parser;
+use service::{init_tracing, WorldClient};
+use std::{net::SocketAddr, time::Duration};
 use tarpc::{client, context, tokio_serde::formats::Json};
+use tokio::time::sleep;
+use tracing::Instrument;
+
+#[derive(Parser)]
+struct Flags {
+    /// Sets the server address to connect to.
+    #[clap(long)]
+    server_addr: SocketAddr,
+    /// Sets the name to say hello to.
+    #[clap(long)]
+    name: String,
+}

 #[tokio::main]
-async fn main() -> io::Result<()> {
-    env_logger::init();
+async fn main() -> anyhow::Result<()> {
+    let flags = Flags::parse();
+    init_tracing("Tarpc Example Client")?;

-    let flags = App::new("Hello Client")
-        .version("0.1")
-        .author("Tim <tikue@google.com>")
-        .about("Say hello!")
-        .arg(
-            Arg::with_name("server_addr")
-                .long("server_addr")
-                .value_name("ADDRESS")
-                .help("Sets the server address to connect to.")
-                .required(true)
-                .takes_value(true),
-        )
-        .arg(
-            Arg::with_name("name")
-                .short("n")
-                .long("name")
-                .value_name("STRING")
-                .help("Sets the name to say hello to.")
-                .required(true)
-                .takes_value(true),
-        )
-        .get_matches();
-
-    let server_addr = flags.value_of("server_addr").unwrap();
-    let server_addr = server_addr
-        .parse::<SocketAddr>()
-        .unwrap_or_else(|e| panic!(r#"--server_addr value "{}" invalid: {}"#, server_addr, e));
-
-    let name = flags.value_of("name").unwrap().into();
-
-    let mut transport = tarpc::serde_transport::tcp::connect(server_addr, Json::default);
-    transport.config_mut().max_frame_length(usize::MAX);
+    let transport = tarpc::serde_transport::tcp::connect(flags.server_addr, Json::default);

    // WorldClient is generated by the service attribute. It has a constructor `new` that takes a
    // config and any Transport as input.
-    let client = service::WorldClient::new(client::Config::default(), transport.await?).spawn()?;
+    let client = WorldClient::new(client::Config::default(), transport.await?).spawn();

-    // The client has an RPC method for each RPC defined in the annotated trait. It takes the same
-    // args as defined, with the addition of a Context, which is always the first arg. The Context
-    // specifies a deadline and trace information which can be helpful in debugging requests.
-    let hello = client.hello(context::current(), name).await?;
+    let hello = async move {
+        // Send the request twice, just to be safe! ;)
+        tokio::select! {
+            hello1 = client.hello(context::current(), format!("{}1", flags.name)) => { hello1 }
+            hello2 = client.hello(context::current(), format!("{}2", flags.name)) => { hello2 }
+        }
+    }
+    .instrument(tracing::info_span!("Two Hellos"))
+    .await;

-    println!("{}", hello);
+    tracing::info!("{:?}", hello);
+
+    // Let the background span processor finish.
+    sleep(Duration::from_micros(1)).await;
+    opentelemetry::global::shutdown_tracer_provider();

    Ok(())
 }
--- a/example-service/src/lib.rs
+++ b/example-service/src/lib.rs
@@ -4,6 +4,9 @@
 // license that can be found in the LICENSE file or at
 // https://opensource.org/licenses/MIT.

+use std::env;
+use tracing_subscriber::{fmt::format::FmtSpan, prelude::*};
+
 /// This is the service definition. It looks a lot like a trait definition.
 /// It defines one RPC, hello, which takes one arg, name, and returns a String.
 #[tarpc::service]
@@ -11,3 +14,21 @@ pub trait World {
    /// Returns a greeting for name.
    async fn hello(name: String) -> String;
 }
+
+/// Initializes an OpenTelemetry tracing subscriber with a Jaeger backend.
+pub fn init_tracing(service_name: &str) -> anyhow::Result<()> {
+    env::set_var("OTEL_BSP_MAX_EXPORT_BATCH_SIZE", "12");
+
+    let tracer = opentelemetry_jaeger::new_pipeline()
+        .with_service_name(service_name)
+        .with_max_packet_size(2usize.pow(13))
+        .install_batch(opentelemetry::runtime::Tokio)?;
+
+    tracing_subscriber::registry()
+        .with(tracing_subscriber::EnvFilter::from_default_env())
+        .with(tracing_subscriber::fmt::layer().with_span_events(FmtSpan::NEW | FmtSpan::CLOSE))
+        .with(tracing_opentelemetry::layer().with_tracer(tracer))
+        .try_init()?;
+
+    Ok(())
+}
--- a/example-service/src/server.rs
+++ b/example-service/src/server.rs
@@ -4,18 +4,30 @@
 // license that can be found in the LICENSE file or at
 // https://opensource.org/licenses/MIT.

-use clap::{App, Arg};
+use clap::Parser;
 use futures::{future, prelude::*};
-use service::World;
+use rand::{
+    distributions::{Distribution, Uniform},
+    thread_rng,
+};
+use service::{init_tracing, World};
 use std::{
-    io,
-    net::{IpAddr, SocketAddr},
+    net::{IpAddr, Ipv6Addr, SocketAddr},
+    time::Duration,
 };
 use tarpc::{
    context,
-    server::{self, Channel, Incoming},
+    server::{self, incoming::Incoming, Channel},
    tokio_serde::formats::Json,
 };
+use tokio::time;
+
+#[derive(Parser)]
+struct Flags {
+    /// Sets the port number to listen on.
+    #[clap(long)]
+    port: u16,
+}

 // This is the type that implements the generated World trait. It is the business logic
 // and is used to start the server.
@@ -25,35 +37,19 @@ struct HelloServer(SocketAddr);
 #[tarpc::server]
 impl World for HelloServer {
    async fn hello(self, _: context::Context, name: String) -> String {
-        format!("Hello, {}! You are connected from {:?}.", name, self.0)
+        let sleep_time =
+            Duration::from_millis(Uniform::new_inclusive(1, 10).sample(&mut thread_rng()));
+        time::sleep(sleep_time).await;
+        format!("Hello, {name}! You are connected from {}", self.0)
    }
 }

 #[tokio::main]
-async fn main() -> io::Result<()> {
-    env_logger::init();
+async fn main() -> anyhow::Result<()> {
+    let flags = Flags::parse();
+    init_tracing("Tarpc Example Server")?;

-    let flags = App::new("Hello Server")
-        .version("0.1")
-        .author("Tim <tikue@google.com>")
-        .about("Say hello!")
-        .arg(
-            Arg::with_name("port")
-                .short("p")
-                .long("port")
-                .value_name("NUMBER")
-                .help("Sets the port number to listen on")
-                .required(true)
-                .takes_value(true),
-        )
-        .get_matches();
-
-    let port = flags.value_of("port").unwrap();
-    let port = port
-        .parse()
-        .unwrap_or_else(|e| panic!(r#"--port value "{}" invalid: {}"#, port, e));
-
-    let server_addr = (IpAddr::from([0, 0, 0, 0]), port);
+    let server_addr = (IpAddr::V6(Ipv6Addr::LOCALHOST), flags.port);

    // JSON transport is provided by the json_transport tarpc module. It makes it easy
    // to start up a serde-powered json serialization strategy over TCP.
@@ -64,12 +60,12 @@ async fn main() -> io::Result<()> {
        .filter_map(|r| future::ready(r.ok()))
        .map(server::BaseChannel::with_defaults)
        // Limit channels to 1 per IP.
-        .max_channels_per_key(1, |t| t.as_ref().peer_addr().unwrap().ip())
+        .max_channels_per_key(1, |t| t.transport().peer_addr().unwrap().ip())
        // serve is generated by the service attribute. It takes as input any type implementing
        // the generated World trait.
        .map(|channel| {
-            let server = HelloServer(channel.as_ref().as_ref().peer_addr().unwrap());
-            channel.requests().execute(server.serve())
+            let server = HelloServer(channel.transport().peer_addr().unwrap());
+            channel.execute(server.serve())
        })
        // Max 10 channels.
        .buffer_unordered(10)
--- a/hooks/pre-commit
+++ b/hooks/pre-commit
@@ -67,7 +67,7 @@ else
 fi

 printf "${PREFIX} Checking for rustfmt ... "
-command -v cargo fmt &>/dev/null
+command -v rustfmt &>/dev/null
 if [ $? == 0 ]; then
 	printf "${SUCCESS}\n"
 else
@@ -93,19 +93,19 @@ diff=""
 for file in $(git diff --name-only --cached);
 do
 	if [ ${file: -3} == ".rs" ]; then
-        diff="$diff$(cargo fmt -- --check $file)"
+        diff="$diff$(rustfmt --edition 2018 --check $file)"
+        if [ $? != 0 ]; then
+            FMTRESULT=1
+        fi
 	fi
 done
-if grep --quiet "^[-+]" <<< "$diff"; then
-    FMTRESULT=1
-fi

 if [ "${TARPC_SKIP_RUSTFMT}" == 1 ]; then
 	printf "${SKIPPED}\n"$?
 elif [ ${FMTRESULT} != 0 ]; then
 	FAILED=1
 	printf "${FAILURE}\n"
-    echo "$diff" | sed 's/Using rustfmt config file.*$/d/'
+    echo "$diff"
 else
 	printf "${SUCCESS}\n"
 fi
--- a/plugins/Cargo.toml
+++ b/plugins/Cargo.toml
@@ -1,8 +1,9 @@
 [package]
 name = "tarpc-plugins"
-version = "0.10.0"
+version = "0.12.0"
+rust-version = "1.56"
 authors = ["Adam Wright <adam.austin.wright@gmail.com>", "Tim Kuehn <timothy.j.kuehn@gmail.com>"]
-edition = "2018"
+edition = "2021"
 license = "MIT"
 documentation = "https://docs.rs/tarpc-plugins"
 homepage = "https://github.com/google/tarpc"
@@ -30,4 +31,4 @@ proc-macro = true
 assert-type-eq = "0.1.0"
 futures = "0.3"
 serde = { version = "1.0", features = ["derive"] }
-tarpc = { path = "../tarpc" }
+tarpc = { path = "../tarpc", features = ["serde1"] }
--- a/plugins/LICENSE
+++ b/plugins/LICENSE
@@ -0,0 +1,9 @@
+The MIT License (MIT)
+
+Copyright 2016 Google Inc. All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
--- a/plugins/src/lib.rs
+++ b/plugins/src/lib.rs
@@ -83,7 +83,7 @@ impl Parse for Service {
                    ident_errors,
                    syn::Error::new(
                        rpc.ident.span(),
-                        format!("method name conflicts with generated fn `{}::serve`", ident)
+                        format!("method name conflicts with generated fn `{ident}::serve`")
                    )
                );
            }
@@ -267,18 +267,25 @@ pub fn service(attr: TokenStream, input: TokenStream) -> TokenStream {
        None
    };

+    let methods = rpcs.iter().map(|rpc| &rpc.ident).collect::<Vec<_>>();
+    let request_names = methods
+        .iter()
+        .map(|m| format!("{ident}.{m}"))
+        .collect::<Vec<_>>();
+
    ServiceGenerator {
        response_fut_name,
        service_ident: ident,
        server_ident: &format_ident!("Serve{}", ident),
-        response_fut_ident: &Ident::new(&response_fut_name, ident.span()),
+        response_fut_ident: &Ident::new(response_fut_name, ident.span()),
        client_ident: &format_ident!("{}Client", ident),
        request_ident: &format_ident!("{}Request", ident),
        response_ident: &format_ident!("{}Response", ident),
        vis,
        args,
        method_attrs: &rpcs.iter().map(|rpc| &*rpc.attrs).collect::<Vec<_>>(),
-        method_idents: &rpcs.iter().map(|rpc| &rpc.ident).collect::<Vec<_>>(),
+        method_idents: &methods,
+        request_names: &*request_names,
        attrs,
        rpcs,
        return_types: &rpcs
@@ -299,7 +306,7 @@ pub fn service(attr: TokenStream, input: TokenStream) -> TokenStream {
            .collect::<Vec<_>>(),
        future_types: &camel_case_fn_names
            .iter()
-            .map(|name| parse_str(&format!("{}Fut", name)).unwrap())
+            .map(|name| parse_str(&format!("{name}Fut")).unwrap())
            .collect::<Vec<_>>(),
        derive_serialize: derive_serialize.as_ref(),
    }
@@ -399,14 +406,10 @@ fn verify_types_were_provided(
 ) -> syn::Result<()> {
    let mut result = Ok(());
    for (method, expected) in expected {
-        if provided
-            .iter()
-            .find(|typedecl| typedecl.ident == expected)
-            .is_none()
-        {
+        if !provided.iter().any(|typedecl| typedecl.ident == expected) {
            let mut e = syn::Error::new(
                span,
-                format!("not all trait items implemented, missing: `{}`", expected),
+                format!("not all trait items implemented, missing: `{expected}`"),
            );
            let fn_span = method.sig.fn_token.span();
            e.extend(syn::Error::new(
@@ -441,6 +444,7 @@ struct ServiceGenerator<'a> {
    camel_case_idents: &'a [Ident],
    future_types: &'a [Type],
    method_idents: &'a [&'a Ident],
+    request_names: &'a [String],
    method_attrs: &'a [&'a [Attribute]],
    args: &'a [&'a [PatType]],
    return_types: &'a [&'a Type],
@@ -475,7 +479,7 @@ impl<'a> ServiceGenerator<'a> {
                    ),
                    output,
                )| {
-                    let ty_doc = format!("The response future returned by {}.", ident);
+                    let ty_doc = format!("The response future returned by [`{service_ident}::{ident}`].");
                    quote! {
                        #[doc = #ty_doc]
                        type #future_type: std::future::Future<Output = #output>;
@@ -524,6 +528,7 @@ impl<'a> ServiceGenerator<'a> {
            camel_case_idents,
            arg_pats,
            method_idents,
+            request_names,
            ..
        } = self;

@@ -534,6 +539,16 @@ impl<'a> ServiceGenerator<'a> {
                type Resp = #response_ident;
                type Fut = #response_fut_ident<S>;

+                fn method(&self, req: &#request_ident) -> Option<&'static str> {
+                    Some(match req {
+                        #(
+                            #request_ident::#camel_case_idents{..} => {
+                                #request_names
+                            }
+                        )*
+                    })
+                }
+
                fn serve(self, ctx: tarpc::context::Context, req: #request_ident) -> Self::Fut {
                    match req {
                        #(
@@ -563,6 +578,7 @@ impl<'a> ServiceGenerator<'a> {

        quote! {
            /// The request sent over the wire from the client to the server.
+            #[allow(missing_docs)]
            #[derive(Debug)]
            #derive_serialize
            #vis enum #request_ident {
@@ -583,6 +599,7 @@ impl<'a> ServiceGenerator<'a> {

        quote! {
            /// The response sent over the wire from the server to the client.
+            #[allow(missing_docs)]
            #[derive(Debug)]
            #derive_serialize
            #vis enum #response_ident {
@@ -603,6 +620,7 @@ impl<'a> ServiceGenerator<'a> {

        quote! {
            /// A future resolving to a server response.
+            #[allow(missing_docs)]
            #vis enum #response_fut_ident<S: #service_ident> {
                #( #camel_case_idents(<S as #service_ident>::#future_types) ),*
            }
@@ -714,6 +732,7 @@ impl<'a> ServiceGenerator<'a> {
            method_attrs,
            vis,
            method_idents,
+            request_names,
            args,
            return_types,
            arg_pats,
@@ -727,9 +746,9 @@ impl<'a> ServiceGenerator<'a> {
                    #[allow(unused)]
                    #( #method_attrs )*
                    #vis fn #method_idents(&self, ctx: tarpc::context::Context, #( #args ),*)
-                        -> impl std::future::Future<Output = std::io::Result<#return_types>> + '_ {
+                        -> impl std::future::Future<Output = Result<#return_types, tarpc::client::RpcError>> + '_ {
                        let request = #request_ident::#camel_case_idents { #( #arg_pats ),* };
-                        let resp = self.0.call(ctx, request);
+                        let resp = self.0.call(ctx, #request_names, request);
                        async move {
                            match resp.await? {
                                #response_ident::#camel_case_idents(msg) => std::result::Result::Ok(msg),
--- a/tarpc/Cargo.toml
+++ b/tarpc/Cargo.toml
@@ -1,8 +1,12 @@
 [package]
 name = "tarpc"
-version = "0.25.1"
-authors = ["Adam Wright <adam.austin.wright@gmail.com>", "Tim Kuehn <timothy.j.kuehn@gmail.com>"]
-edition = "2018"
+version = "0.30.0"
+rust-version = "1.58.0"
+authors = [
+    "Adam Wright <adam.austin.wright@gmail.com>",
+    "Tim Kuehn <timothy.j.kuehn@gmail.com>",
+]
+edition = "2021"
 license = "MIT"
 documentation = "https://docs.rs/tarpc"
 homepage = "https://github.com/google/tarpc"
@@ -16,11 +20,20 @@ description = "An RPC framework for Rust with a focus on ease of use."
 default = []

 serde1 = ["tarpc-plugins/serde1", "serde", "serde/derive"]
-tokio1 = ["tokio/rt-multi-thread"]
-serde-transport = ["serde1", "tokio1", "tokio-serde/json", "tokio-util/codec"]
+tokio1 = ["tokio/rt"]
+serde-transport = ["serde1", "tokio1", "tokio-serde", "tokio-util/codec"]
+serde-transport-json = ["tokio-serde/json"]
+serde-transport-bincode = ["tokio-serde/bincode"]
 tcp = ["tokio/net"]

-full = ["serde1", "tokio1", "serde-transport", "tcp"]
+full = [
+    "serde1",
+    "tokio1",
+    "serde-transport",
+    "serde-transport-json",
+    "serde-transport-bincode",
+    "tcp",
+]

 [badges]
 travis-ci = { repository = "google/tarpc" }
@@ -30,26 +43,36 @@ anyhow = "1.0"
 fnv = "1.0"
 futures = "0.3"
 humantime = "2.0"
-log = "0.4"
 pin-project = "1.0"
-rand = "0.7"
+rand = "0.8"
 serde = { optional = true, version = "1.0", features = ["derive"] }
 static_assertions = "1.1.0"
-tarpc-plugins = { path = "../plugins", version = "0.10" }
+tarpc-plugins = { path = "../plugins", version = "0.12" }
+thiserror = "1.0"
 tokio = { version = "1", features = ["time"] }
-tokio-util = { version = "0.6.3", features = ["time"] }
+tokio-util = { version = "0.7.3", features = ["time"] }
 tokio-serde = { optional = true, version = "0.8" }
+tracing = { version = "0.1", default-features = false, features = [
+    "attributes",
+    "log",
+] }
+tracing-opentelemetry = { version = "0.17.2", default-features = false }
+opentelemetry = { version = "0.17.0", default-features = false }
+

 [dev-dependencies]
 assert_matches = "1.4"
 bincode = "1.3"
 bytes = { version = "1", features = ["serde"] }
-env_logger = "0.8"
 flate2 = "1.0"
 futures-test = "0.3"
-log = "0.4"
+opentelemetry = { version = "0.17.0", default-features = false, features = [
+    "rt-tokio",
+] }
+opentelemetry-jaeger = { version = "0.16.0", features = ["rt-tokio"] }
 pin-utils = "0.1.0-alpha"
 serde_bytes = "0.11"
+tracing-subscriber = { version = "0.3", features = ["env-filter"] }
 tokio = { version = "1", features = ["full", "test-util"] }
 tokio-serde = { version = "0.8", features = ["json", "bincode"] }
 trybuild = "1.0"
@@ -63,7 +86,7 @@ name = "compression"
 required-features = ["serde-transport", "tcp"]

 [[example]]
-name = "server_calling_server"
+name = "tracing"
 required-features = ["full"]

 [[example]]
--- a/tarpc/LICENSE
+++ b/tarpc/LICENSE
@@ -0,0 +1,9 @@
+The MIT License (MIT)
+
+Copyright 2016 Google Inc. All Rights Reserved.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
--- a/tarpc/examples/compression.rs
+++ b/tarpc/examples/compression.rs
@@ -7,8 +7,8 @@ use tarpc::{
    client, context,
    serde_transport::tcp,
    server::{BaseChannel, Channel},
+    tokio_serde::formats::Bincode,
 };
-use tokio_serde::formats::Bincode;

 /// Type of compression that should be enabled on the request. The transport is free to ignore this.
 #[derive(Debug, PartialEq, Eq, Clone, Copy, Deserialize, Serialize)]
@@ -54,7 +54,7 @@ where
            if algorithm != CompressionAlgorithm::Deflate {
                return Err(io::Error::new(
                    io::ErrorKind::InvalidData,
-                    format!("Compression algorithm {:?} not supported", algorithm),
+                    format!("Compression algorithm {algorithm:?} not supported"),
                ));
            }
            let mut deflater = DeflateDecoder::new(payload.as_slice());
@@ -102,7 +102,7 @@ struct HelloServer;
 #[tarpc::server]
 impl World for HelloServer {
    async fn hello(self, _: context::Context, name: String) -> String {
-        format!("Hey, {}!", name)
+        format!("Hey, {name}!")
    }
 }

@@ -118,7 +118,7 @@ async fn main() -> anyhow::Result<()> {
    });

    let transport = tcp::connect(addr, Bincode::default).await?;
-    let client = WorldClient::new(client::Config::default(), add_compression(transport)).spawn()?;
+    let client = WorldClient::new(client::Config::default(), add_compression(transport)).spawn();

    println!(
        "{}",
--- a/tarpc/examples/custom_transport.rs
+++ b/tarpc/examples/custom_transport.rs
@@ -1,10 +1,9 @@
-use futures::future;
 use tarpc::context::Context;
 use tarpc::serde_transport as transport;
 use tarpc::server::{BaseChannel, Channel};
+use tarpc::tokio_serde::formats::Bincode;
+use tarpc::tokio_util::codec::length_delimited::LengthDelimitedCodec;
 use tokio::net::{UnixListener, UnixStream};
-use tokio_serde::formats::Bincode;
-use tokio_util::codec::length_delimited::LengthDelimitedCodec;

 #[tarpc::service]
 pub trait PingService {
@@ -14,16 +13,13 @@ pub trait PingService {
 #[derive(Clone)]
 struct Service;

+#[tarpc::server]
 impl PingService for Service {
-    type PingFut = future::Ready<()>;
-
-    fn ping(self, _: Context) -> Self::PingFut {
-        future::ready(())
-    }
+    async fn ping(self, _: Context) {}
 }

 #[tokio::main]
-async fn main() -> std::io::Result<()> {
+async fn main() -> anyhow::Result<()> {
    let bind_addr = "/tmp/tarpc_on_unix_example.sock";

    let _ = std::fs::remove_file(bind_addr);
@@ -44,7 +40,9 @@ async fn main() -> std::io::Result<()> {
    let conn = UnixStream::connect(bind_addr).await?;
    let transport = transport::new(codec_builder.new_framed(conn), Bincode::default());
    PingServiceClient::new(Default::default(), transport)
-        .spawn()?
+        .spawn()
        .ping(tarpc::context::current())
-        .await
+        .await?;
+
+    Ok(())
 }
--- a/tarpc/examples/pubsub.rs
+++ b/tarpc/examples/pubsub.rs
@@ -38,10 +38,11 @@ use futures::{
    future::{self, AbortHandle},
    prelude::*,
 };
-use log::info;
 use publisher::Publisher as _;
 use std::{
    collections::HashMap,
+    env,
+    error::Error,
    io,
    net::SocketAddr,
    sync::{Arc, Mutex, RwLock},
@@ -51,9 +52,11 @@ use tarpc::{
    client, context,
    serde_transport::tcp,
    server::{self, Channel},
+    tokio_serde::formats::Json,
 };
 use tokio::net::ToSocketAddrs;
-use tokio_serde::formats::Json;
+use tracing::info;
+use tracing_subscriber::prelude::*;

 pub mod subscriber {
    #[tarpc::service]
@@ -83,10 +86,7 @@ impl subscriber::Subscriber for Subscriber {
    }

    async fn receive(self, _: context::Context, topic: String, message: String) {
-        info!(
-            "[{}] received message on topic '{}': {}",
-            self.local_addr, topic, message
-        );
+        info!(local_addr = %self.local_addr, %topic, %message, "ReceivedMessage")
    }
 }

@@ -120,7 +120,7 @@ impl Subscriber {
        let (handler, abort_handle) = future::abortable(handler.execute(subscriber.serve()));
        tokio::spawn(async move {
            match handler.await {
-                Ok(()) | Err(future::Aborted) => info!("[{}] subscriber shutdown.", local_addr),
+                Ok(()) | Err(future::Aborted) => info!(?local_addr, "subscriber shutdown."),
            }
        });
        Ok(SubscriberHandle(abort_handle))
@@ -129,7 +129,6 @@ impl Subscriber {

 #[derive(Debug)]
 struct Subscription {
-    subscriber: subscriber::SubscriberClient,
    topics: Vec<String>,
 }

@@ -153,13 +152,13 @@ impl Publisher {
            subscriptions: self.clone().start_subscription_manager().await?,
        };

-        info!("[{}] listening for publishers.", publisher_addrs.publisher);
+        info!(publisher_addr = %publisher_addrs.publisher, "listening for publishers.",);
        tokio::spawn(async move {
            // Because this is just an example, we know there will only be one publisher. In more
            // realistic code, this would be a loop to continually accept new publisher
            // connections.
            let publisher = connecting_publishers.next().await.unwrap().unwrap();
-            info!("[{}] publisher connected.", publisher.peer_addr().unwrap());
+            info!(publisher.peer_addr = ?publisher.peer_addr(), "publisher connected.");

            server::BaseChannel::with_defaults(publisher)
                .execute(self.serve())
@@ -174,7 +173,7 @@ impl Publisher {
            .await?
            .filter_map(|r| future::ready(r.ok()));
        let new_subscriber_addr = connecting_subscribers.get_ref().local_addr();
-        info!("[{}] listening for subscribers.", new_subscriber_addr);
+        info!(?new_subscriber_addr, "listening for subscribers.");

        tokio::spawn(async move {
            while let Some(conn) = connecting_subscribers.next().await {
@@ -210,12 +209,11 @@ impl Publisher {
            self.clients.lock().unwrap().insert(
                subscriber_addr,
                Subscription {
-                    subscriber: subscriber.clone(),
                    topics: topics.clone(),
                },
            );

-            info!("[{}] subscribed to topics: {:?}", subscriber_addr, topics);
+            info!(%subscriber_addr, ?topics, "subscribed to new topics");
            let mut subscriptions = self.subscriptions.write().unwrap();
            for topic in topics {
                subscriptions
@@ -226,18 +224,18 @@ impl Publisher {
        }
    }

-    fn start_subscriber_gc(
+    fn start_subscriber_gc<E: Error>(
        self,
        subscriber_addr: SocketAddr,
-        client_dispatch: impl Future<Output = anyhow::Result<()>> + Send + 'static,
+        client_dispatch: impl Future<Output = Result<(), E>> + Send + 'static,
        subscriber_ready: oneshot::Receiver<()>,
    ) {
        tokio::spawn(async move {
            if let Err(e) = client_dispatch.await {
                info!(
-                    "[{}] subscriber connection broken: {:?}",
-                    subscriber_addr, e
-                )
+                    %subscriber_addr,
+                    error = %e,
+                    "subscriber connection broken");
            }
            // Don't clean up the subscriber until initialization is done.
            let _ = subscriber_ready.await;
@@ -281,13 +279,29 @@ impl publisher::Publisher for Publisher {
    }
 }

+/// Initializes an OpenTelemetry tracing subscriber with a Jaeger backend.
+fn init_tracing(service_name: &str) -> anyhow::Result<()> {
+    env::set_var("OTEL_BSP_MAX_EXPORT_BATCH_SIZE", "12");
+    let tracer = opentelemetry_jaeger::new_pipeline()
+        .with_service_name(service_name)
+        .with_max_packet_size(2usize.pow(13))
+        .install_batch(opentelemetry::runtime::Tokio)?;
+
+    tracing_subscriber::registry()
+        .with(tracing_subscriber::filter::EnvFilter::from_default_env())
+        .with(tracing_subscriber::fmt::layer())
+        .with(tracing_opentelemetry::layer().with_tracer(tracer))
+        .try_init()?;
+
+    Ok(())
+}
+
 #[tokio::main]
 async fn main() -> anyhow::Result<()> {
-    env_logger::init();
+    init_tracing("Pub/Sub")?;

-    let clients = Arc::new(Mutex::new(HashMap::new()));
    let addrs = Publisher {
-        clients,
+        clients: Arc::new(Mutex::new(HashMap::new())),
        subscriptions: Arc::new(RwLock::new(HashMap::new())),
    }
    .start()
@@ -309,7 +323,7 @@ async fn main() -> anyhow::Result<()> {
        client::Config::default(),
        tcp::connect(addrs.publisher, Json::default).await?,
    )
-    .spawn()?;
+    .spawn();

    publisher
        .publish(context::current(), "calculus".into(), "sqrt(2)".into())
@@ -337,6 +351,7 @@ async fn main() -> anyhow::Result<()> {
        )
        .await?;

+    opentelemetry::global::shutdown_tracer_provider();
    info!("done.");

    Ok(())
--- a/tarpc/examples/readme.rs
+++ b/tarpc/examples/readme.rs
@@ -5,7 +5,6 @@
 // https://opensource.org/licenses/MIT.

 use futures::future::{self, Ready};
-use std::io;
 use tarpc::{
    client, context,
    server::{self, Channel},
@@ -30,12 +29,12 @@ impl World for HelloServer {
    type HelloFut = Ready<String>;

    fn hello(self, _: context::Context, name: String) -> Self::HelloFut {
-        future::ready(format!("Hello, {}!", name))
+        future::ready(format!("Hello, {name}!"))
    }
 }

 #[tokio::main]
-async fn main() -> io::Result<()> {
+async fn main() -> anyhow::Result<()> {
    let (client_transport, server_transport) = tarpc::transport::channel::unbounded();

    let server = server::BaseChannel::with_defaults(server_transport);
@@ -43,14 +42,14 @@ async fn main() -> io::Result<()> {

    // WorldClient is generated by the #[tarpc::service] attribute. It has a constructor `new`
    // that takes a config and any Transport as input.
-    let client = WorldClient::new(client::Config::default(), client_transport).spawn()?;
+    let client = WorldClient::new(client::Config::default(), client_transport).spawn();

    // The client has an RPC method for each RPC defined in the annotated trait. It takes the same
    // args as defined, with the addition of a Context, which is always the first arg. The Context
    // specifies a deadline and trace information which can be helpful in debugging requests.
    let hello = client.hello(context::current(), "Stim".to_string()).await?;

-    println!("{}", hello);
+    println!("{hello}");

    Ok(())
 }
--- a/tarpc/examples/server_calling_server.rs
+++ b/tarpc/examples/server_calling_server.rs
@@ -6,12 +6,12 @@

 use crate::{add::Add as AddService, double::Double as DoubleService};
 use futures::{future, prelude::*};
-use std::io;
 use tarpc::{
    client, context,
-    server::{BaseChannel, Incoming},
+    server::{incoming::Incoming, BaseChannel},
+    tokio_serde::formats::Json,
 };
-use tokio_serde::formats::Json;
+use tracing_subscriber::prelude::*;

 pub mod add {
    #[tarpc::service]
@@ -54,9 +54,25 @@ impl DoubleService for DoubleServer {
    }
 }

+fn init_tracing(service_name: &str) -> anyhow::Result<()> {
+    let tracer = opentelemetry_jaeger::new_pipeline()
+        .with_service_name(service_name)
+        .with_auto_split_batch(true)
+        .with_max_packet_size(2usize.pow(13))
+        .install_batch(opentelemetry::runtime::Tokio)?;
+
+    tracing_subscriber::registry()
+        .with(tracing_subscriber::EnvFilter::from_default_env())
+        .with(tracing_subscriber::fmt::layer())
+        .with(tracing_opentelemetry::layer().with_tracer(tracer))
+        .try_init()?;
+
+    Ok(())
+}
+
 #[tokio::main]
-async fn main() -> io::Result<()> {
-    env_logger::init();
+async fn main() -> anyhow::Result<()> {
+    init_tracing("tarpc_tracing_example")?;

    let add_listener = tarpc::serde_transport::tcp::listen("localhost:0", Json::default)
        .await?
@@ -69,7 +85,7 @@ async fn main() -> io::Result<()> {
    tokio::spawn(add_server);

    let to_add_server = tarpc::serde_transport::tcp::connect(addr, Json::default).await?;
-    let add_client = add::AddClient::new(client::Config::default(), to_add_server).spawn()?;
+    let add_client = add::AddClient::new(client::Config::default(), to_add_server).spawn();

    let double_listener = tarpc::serde_transport::tcp::listen("localhost:0", Json::default)
        .await?
@@ -83,10 +99,14 @@ async fn main() -> io::Result<()> {

    let to_double_server = tarpc::serde_transport::tcp::connect(addr, Json::default).await?;
    let double_client =
-        double::DoubleClient::new(client::Config::default(), to_double_server).spawn()?;
+        double::DoubleClient::new(client::Config::default(), to_double_server).spawn();

-    for i in 1..=5 {
-        eprintln!("{:?}", double_client.double(context::current(), i).await?);
+    let ctx = context::current();
+    for _ in 1..=5 {
+        tracing::info!("{:?}", double_client.double(ctx, 1).await?);
    }
+
+    opentelemetry::global::shutdown_tracer_provider();
+
    Ok(())
 }
--- a/tarpc/src/cancellations.rs
+++ b/tarpc/src/cancellations.rs
@@ -0,0 +1,49 @@
+use futures::{prelude::*, task::*};
+use std::pin::Pin;
+use tokio::sync::mpsc;
+
+/// Sends request cancellation signals.
+#[derive(Debug, Clone)]
+pub struct RequestCancellation(mpsc::UnboundedSender<u64>);
+
+/// A stream of IDs of requests that have been canceled.
+#[derive(Debug)]
+pub struct CanceledRequests(mpsc::UnboundedReceiver<u64>);
+
+/// Returns a channel to send request cancellation messages.
+pub fn cancellations() -> (RequestCancellation, CanceledRequests) {
+    // Unbounded because messages are sent in the drop fn. This is fine, because it's still
+    // bounded by the number of in-flight requests.
+    let (tx, rx) = mpsc::unbounded_channel();
+    (RequestCancellation(tx), CanceledRequests(rx))
+}
+
+impl RequestCancellation {
+    /// Cancels the request with ID `request_id`.
+    ///
+    /// No validation is done of `request_id`. There is no way to know if the request id provided
+    /// corresponds to a request actually tracked by the backing channel. `RequestCancellation` is
+    /// a one-way communication channel.
+    ///
+    /// Once request data is cleaned up, a response will never be received by the client. This is
+    /// useful primarily when request processing ends prematurely for requests with long deadlines
+    /// which would otherwise continue to be tracked by the backing channel—a kind of leak.
+    pub fn cancel(&self, request_id: u64) {
+        let _ = self.0.send(request_id);
+    }
+}
+
+impl CanceledRequests {
+    /// Polls for a cancelled request.
+    pub fn poll_recv(&mut self, cx: &mut Context<'_>) -> Poll<Option<u64>> {
+        self.0.poll_recv(cx)
+    }
+}
+
+impl Stream for CanceledRequests {
+    type Item = u64;
+
+    fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<u64>> {
+        self.poll_recv(cx)
+    }
+}
--- a/tarpc/src/client.rs
+++ b/tarpc/src/client.rs
@@ -9,15 +9,16 @@
 mod in_flight_requests;

 use crate::{
-    context, trace::SpanId, ClientMessage, PollContext, PollIo, Request, Response, Transport,
+    cancellations::{cancellations, CanceledRequests, RequestCancellation},
+    context, trace, ClientMessage, Request, Response, ServerError, Transport,
 };
 use futures::{prelude::*, ready, stream::Fuse, task::*};
-use in_flight_requests::InFlightRequests;
-use log::{info, trace};
+use in_flight_requests::{DeadlineExceededError, InFlightRequests};
 use pin_project::pin_project;
 use std::{
    convert::TryFrom,
-    fmt, io,
+    error::Error,
+    fmt, mem,
    pin::Pin,
    sync::{
        atomic::{AtomicUsize, Ordering},
@@ -25,6 +26,7 @@ use std::{
    },
 };
 use tokio::sync::{mpsc, oneshot};
+use tracing::Span;

 /// Settings that control the behavior of the client.
 #[derive(Clone, Debug)]
@@ -61,19 +63,18 @@ pub struct NewClient<C, D> {
 impl<C, D, E> NewClient<C, D>
 where
    D: Future<Output = Result<(), E>> + Send + 'static,
-    E: std::fmt::Display,
+    E: std::error::Error + Send + Sync + 'static,
 {
    /// Helper method to spawn the dispatch on the default executor.
    #[cfg(feature = "tokio1")]
    #[cfg_attr(docsrs, doc(cfg(feature = "tokio1")))]
-    pub fn spawn(self) -> io::Result<C> {
-        use log::warn;
-
-        let dispatch = self
-            .dispatch
-            .unwrap_or_else(move |e| warn!("Connection broken: {}", e));
+    pub fn spawn(self) -> C {
+        let dispatch = self.dispatch.unwrap_or_else(move |e| {
+            let e = anyhow::Error::new(e);
+            tracing::warn!("Connection broken: {:?}", e);
+        });
        tokio::spawn(dispatch);
-        Ok(self.client)
+        self.client
    }
 }

@@ -83,14 +84,10 @@ impl<C, D> fmt::Debug for NewClient<C, D> {
    }
 }

-#[allow(dead_code)]
-#[allow(clippy::no_effect)]
-const CHECK_USIZE: () = {
-    if std::mem::size_of::<usize>() > std::mem::size_of::<u64>() {
-        // TODO: replace this with panic!() as soon as RFC 2345 gets stabilized
-        ["usize is too big to fit in u64"][42];
-    }
-};
+const _CHECK_USIZE: () = assert!(
+    std::mem::size_of::<usize>() <= std::mem::size_of::<u64>(),
+    "usize is too big to fit in u64"
+);

 /// Handles communication from the client to request dispatch.
 #[derive(Debug)]
@@ -114,100 +111,119 @@ impl<Req, Resp> Clone for Channel<Req, Resp> {

 impl<Req, Resp> Channel<Req, Resp> {
    /// Sends a request to the dispatch task to forward to the server, returning a [`Future`] that
-    /// resolves when the request is sent (not when the response is received).
-    fn send(
+    /// resolves to the response.
+    #[tracing::instrument(
+        name = "RPC",
+        skip(self, ctx, request_name, request),
+        fields(
+            rpc.trace_id = tracing::field::Empty,
+            rpc.deadline = %humantime::format_rfc3339(ctx.deadline),
+            otel.kind = "client",
+            otel.name = request_name)
+        )]
+    pub async fn call(
        &self,
        mut ctx: context::Context,
+        request_name: &str,
        request: Req,
-    ) -> impl Future<Output = io::Result<DispatchResponse<Resp>>> + '_ {
-        // Convert the context to the call context.
-        ctx.trace_context.parent_id = Some(ctx.trace_context.span_id);
-        ctx.trace_context.span_id = SpanId::random(&mut rand::thread_rng());
-
-        let (response_completion, response) = oneshot::channel();
-        let cancellation = self.cancellation.clone();
+    ) -> Result<Resp, RpcError> {
+        let span = Span::current();
+        ctx.trace_context = trace::Context::try_from(&span).unwrap_or_else(|_| {
+            tracing::trace!(
+                "OpenTelemetry subscriber not installed; making unsampled child context."
+            );
+            ctx.trace_context.new_child()
+        });
+        span.record("rpc.trace_id", &tracing::field::display(ctx.trace_id()));
+        let (response_completion, mut response) = oneshot::channel();
        let request_id =
            u64::try_from(self.next_request_id.fetch_add(1, Ordering::Relaxed)).unwrap();

-        // DispatchResponse impls Drop to cancel in-flight requests. It should be created before
+        // ResponseGuard impls Drop to cancel in-flight requests. It should be created before
        // sending out the request; otherwise, the response future could be dropped after the
-        // request is sent out but before DispatchResponse is created, rendering the cancellation
+        // request is sent out but before ResponseGuard is created, rendering the cancellation
        // logic inactive.
-        let response = DispatchResponse {
-            response,
+        let response_guard = ResponseGuard {
+            response: &mut response,
            request_id,
-            cancellation: Some(cancellation),
-            ctx,
+            cancellation: &self.cancellation,
        };
-        async move {
-            self.to_dispatch
-                .send(DispatchRequest {
-                    ctx,
-                    request_id,
-                    request,
-                    response_completion,
-                })
-                .await
-                .map_err(|mpsc::error::SendError(_)| {
-                    io::Error::from(io::ErrorKind::ConnectionReset)
-                })?;
-            Ok(response)
-        }
-    }
-
-    /// Sends a request to the dispatch task to forward to the server, returning a [`Future`] that
-    /// resolves to the response.
-    pub async fn call(&self, ctx: context::Context, request: Req) -> io::Result<Resp> {
-        let dispatch_response = self.send(ctx, request).await?;
-        dispatch_response.await
+        self.to_dispatch
+            .send(DispatchRequest {
+                ctx,
+                span,
+                request_id,
+                request,
+                response_completion,
+            })
+            .await
+            .map_err(|mpsc::error::SendError(_)| RpcError::Disconnected)?;
+        response_guard.response().await
    }
 }

 /// A server response that is completed by request dispatch when the corresponding response
 /// arrives off the wire.
-#[derive(Debug)]
-struct DispatchResponse<Resp> {
-    response: oneshot::Receiver<Response<Resp>>,
-    ctx: context::Context,
-    cancellation: Option<RequestCancellation>,
+struct ResponseGuard<'a, Resp> {
+    response: &'a mut oneshot::Receiver<Result<Response<Resp>, DeadlineExceededError>>,
+    cancellation: &'a RequestCancellation,
    request_id: u64,
 }

-impl<Resp> Future for DispatchResponse<Resp> {
-    type Output = io::Result<Resp>;
+/// An error that can occur in the processing of an RPC. This is not request-specific errors but
+/// rather cross-cutting errors that can always occur.
+#[derive(thiserror::Error, Clone, Debug, PartialEq, Eq, Hash)]
+#[cfg_attr(feature = "serde1", derive(serde::Serialize, serde::Deserialize))]
+pub enum RpcError {
+    /// The client disconnected from the server.
+    #[error("the client disconnected from the server")]
+    Disconnected,
+    /// The request exceeded its deadline.
+    #[error("the request exceeded its deadline")]
+    DeadlineExceeded,
+    /// The server aborted request processing.
+    #[error("the server aborted request processing")]
+    Server(#[from] ServerError),
+}

-    fn poll(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<Resp>> {
-        let resp = ready!(self.response.poll_unpin(cx));
-        self.cancellation.take();
-        Poll::Ready(match resp {
-            Ok(resp) => Ok(resp.message?),
+impl From<DeadlineExceededError> for RpcError {
+    fn from(_: DeadlineExceededError) -> Self {
+        RpcError::DeadlineExceeded
+    }
+}
+
+impl<Resp> ResponseGuard<'_, Resp> {
+    async fn response(mut self) -> Result<Resp, RpcError> {
+        let response = (&mut self.response).await;
+        // Cancel drop logic once a response has been received.
+        mem::forget(self);
+        match response {
+            Ok(resp) => Ok(resp?.message?),
            Err(oneshot::error::RecvError { .. }) => {
                // The oneshot is Canceled when the dispatch task ends. In that case,
                // there's nothing listening on the other side, so there's no point in
                // propagating cancellation.
-                Err(io::Error::from(io::ErrorKind::ConnectionReset))
+                Err(RpcError::Disconnected)
            }
-        })
+        }
    }
 }

 // Cancels the request when dropped, if not already complete.
-impl<Resp> Drop for DispatchResponse<Resp> {
+impl<Resp> Drop for ResponseGuard<'_, Resp> {
    fn drop(&mut self) {
-        if let Some(cancellation) = &mut self.cancellation {
-            // The receiver needs to be closed to handle the edge case that the request has not
-            // yet been received by the dispatch task. It is possible for the cancel message to
-            // arrive before the request itself, in which case the request could get stuck in the
-            // dispatch map forever if the server never responds (e.g. if the server dies while
-            // responding). Even if the server does respond, it will have unnecessarily done work
-            // for a client no longer waiting for a response. To avoid this, the dispatch task
-            // checks if the receiver is closed before inserting the request in the map. By
-            // closing the receiver before sending the cancel message, it is guaranteed that if the
-            // dispatch task misses an early-arriving cancellation message, then it will see the
-            // receiver as closed.
-            self.response.close();
-            cancellation.cancel(self.request_id);
-        }
+        // The receiver needs to be closed to handle the edge case that the request has not
+        // yet been received by the dispatch task. It is possible for the cancel message to
+        // arrive before the request itself, in which case the request could get stuck in the
+        // dispatch map forever if the server never responds (e.g. if the server dies while
+        // responding). Even if the server does respond, it will have unnecessarily done work
+        // for a client no longer waiting for a response. To avoid this, the dispatch task
+        // checks if the receiver is closed before inserting the request in the map. By
+        // closing the receiver before sending the cancel message, it is guaranteed that if the
+        // dispatch task misses an early-arriving cancellation message, then it will see the
+        // receiver as closed.
+        self.response.close();
+        self.cancellation.cancel(self.request_id);
    }
 }

@@ -242,6 +258,7 @@ where

 /// Handles the lifecycle of requests, writing requests to the wire, managing cancellations,
 /// and dispatching responses to the appropriate channel.
+#[must_use]
 #[pin_project]
 #[derive(Debug)]
 pub struct RequestDispatch<Req, Resp, C> {
@@ -258,6 +275,32 @@ pub struct RequestDispatch<Req, Resp, C> {
    config: Config,
 }

+/// Critical errors that result in a Channel disconnecting.
+#[derive(thiserror::Error, Debug)]
+pub enum ChannelError<E>
+where
+    E: Error + Send + Sync + 'static,
+{
+    /// Could not read from the transport.
+    #[error("could not read from the transport")]
+    Read(#[source] E),
+    /// Could not ready the transport for writes.
+    #[error("could not ready the transport for writes")]
+    Ready(#[source] E),
+    /// Could not write to the transport.
+    #[error("could not write to the transport")]
+    Write(#[source] E),
+    /// Could not flush the transport.
+    #[error("could not flush the transport")]
+    Flush(#[source] E),
+    /// Could not close the write end of the transport.
+    #[error("could not close the write end of the transport")]
+    Close(#[source] E),
+    /// Could not poll expired requests.
+    #[error("could not poll expired requests")]
+    Timer(#[source] tokio::time::error::Error),
+}
+
 impl<Req, Resp, C> RequestDispatch<Req, Resp, C>
 where
    C: Transport<ClientMessage<Req>, Response<Resp>>,
@@ -270,6 +313,42 @@ where
        self.as_mut().project().transport
    }

+    fn poll_ready<'a>(
+        self: &'a mut Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Result<(), ChannelError<C::Error>>> {
+        self.transport_pin_mut()
+            .poll_ready(cx)
+            .map_err(ChannelError::Ready)
+    }
+
+    fn start_send(
+        self: &mut Pin<&mut Self>,
+        message: ClientMessage<Req>,
+    ) -> Result<(), ChannelError<C::Error>> {
+        self.transport_pin_mut()
+            .start_send(message)
+            .map_err(ChannelError::Write)
+    }
+
+    fn poll_flush<'a>(
+        self: &'a mut Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Result<(), ChannelError<C::Error>>> {
+        self.transport_pin_mut()
+            .poll_flush(cx)
+            .map_err(ChannelError::Flush)
+    }
+
+    fn poll_close<'a>(
+        self: &'a mut Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Result<(), ChannelError<C::Error>>> {
+        self.transport_pin_mut()
+            .poll_close(cx)
+            .map_err(ChannelError::Close)
+    }
+
    fn canceled_requests_mut<'a>(self: &'a mut Pin<&mut Self>) -> &'a mut CanceledRequests {
        self.as_mut().project().canceled_requests
    }
@@ -280,17 +359,22 @@ where
        self.as_mut().project().pending_requests
    }

-    fn pump_read(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> PollIo<()> {
-        Poll::Ready(match ready!(self.transport_pin_mut().poll_next(cx)?) {
-            Some(response) => {
+    fn pump_read(
+        mut self: Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Option<Result<(), ChannelError<C::Error>>>> {
+        self.transport_pin_mut()
+            .poll_next(cx)
+            .map_err(ChannelError::Read)
+            .map_ok(|response| {
                self.complete(response);
-                Some(Ok(()))
-            }
-            None => None,
-        })
+            })
    }

-    fn pump_write(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> PollIo<()> {
+    fn pump_write(
+        mut self: Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Option<Result<(), ChannelError<C::Error>>>> {
        enum ReceiverStatus {
            Pending,
            Closed,
@@ -311,7 +395,7 @@ where
        // Receiving Poll::Ready(None) when polling expired requests never indicates "Closed",
        // because there can temporarily be zero in-flight rquests. Therefore, there is no need to
        // track the status like is done with pending and cancelled requests.
-        if let Poll::Ready(Some(_)) = self.in_flight_requests().poll_expired(cx)? {
+        if let Poll::Ready(Some(_)) = self.in_flight_requests().poll_expired(cx) {
            // Expired requests are considered complete; there is no compelling reason to send a
            // cancellation message to the server, since it will have already exhausted its
            // allotted processing time.
@@ -320,12 +404,12 @@ where

        match (pending_requests_status, canceled_requests_status) {
            (ReceiverStatus::Closed, ReceiverStatus::Closed) => {
-                ready!(self.transport_pin_mut().poll_flush(cx)?);
+                ready!(self.poll_close(cx)?);
                Poll::Ready(None)
            }
            (ReceiverStatus::Pending, _) | (_, ReceiverStatus::Pending) => {
                // No more messages to process, so flush any messages buffered in the transport.
-                ready!(self.transport_pin_mut().poll_flush(cx)?);
+                ready!(self.poll_flush(cx)?);

                // Even if we fully-flush, we return Pending, because we have no more requests
                // or cancellations right now.
@@ -341,9 +425,9 @@ where
    fn poll_next_request(
        mut self: Pin<&mut Self>,
        cx: &mut Context<'_>,
-    ) -> PollIo<DispatchRequest<Req, Resp>> {
+    ) -> Poll<Option<Result<DispatchRequest<Req, Resp>, ChannelError<C::Error>>>> {
        if self.in_flight_requests().len() >= self.config.max_in_flight_requests {
-            info!(
+            tracing::info!(
                "At in-flight request capacity ({}/{}).",
                self.in_flight_requests().len(),
                self.config.max_in_flight_requests
@@ -360,10 +444,8 @@ where
            match ready!(self.pending_requests_mut().poll_recv(cx)) {
                Some(request) => {
                    if request.response_completion.is_closed() {
-                        trace!(
-                            "[{}] Request canceled before being sent.",
-                            request.ctx.trace_id()
-                        );
+                        let _entered = request.span.enter();
+                        tracing::info!("AbortRequest");
                        continue;
                    }

@@ -381,14 +463,15 @@ where
    fn poll_next_cancellation(
        mut self: Pin<&mut Self>,
        cx: &mut Context<'_>,
-    ) -> PollIo<(context::Context, u64)> {
+    ) -> Poll<Option<Result<(context::Context, Span, u64), ChannelError<C::Error>>>> {
        ready!(self.ensure_writeable(cx)?);

        loop {
            match ready!(self.canceled_requests_mut().poll_next_unpin(cx)) {
                Some(request_id) => {
-                    if let Some(ctx) = self.in_flight_requests().cancel_request(request_id) {
-                        return Poll::Ready(Some(Ok((ctx, request_id))));
+                    if let Some((ctx, span)) = self.in_flight_requests().cancel_request(request_id)
+                    {
+                        return Poll::Ready(Some(Ok((ctx, span, request_id))));
                    }
                }
                None => return Poll::Ready(None),
@@ -399,54 +482,69 @@ where
    /// Returns Ready if writing a message to the transport (i.e. via write_request or
    /// write_cancel) would not fail due to a full buffer. If the transport is not ready to be
    /// written to, flushes it until it is ready.
-    fn ensure_writeable<'a>(self: &'a mut Pin<&mut Self>, cx: &mut Context<'_>) -> PollIo<()> {
-        while self.transport_pin_mut().poll_ready(cx)?.is_pending() {
-            ready!(self.transport_pin_mut().poll_flush(cx)?);
+    fn ensure_writeable<'a>(
+        self: &'a mut Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Option<Result<(), ChannelError<C::Error>>>> {
+        while self.poll_ready(cx)?.is_pending() {
+            ready!(self.poll_flush(cx)?);
        }
        Poll::Ready(Some(Ok(())))
    }

-    fn poll_write_request<'a>(self: &'a mut Pin<&mut Self>, cx: &mut Context<'_>) -> PollIo<()> {
-        let dispatch_request = match ready!(self.as_mut().poll_next_request(cx)?) {
+    fn poll_write_request<'a>(
+        self: &'a mut Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Option<Result<(), ChannelError<C::Error>>>> {
+        let DispatchRequest {
+            ctx,
+            span,
+            request_id,
+            request,
+            response_completion,
+        } = match ready!(self.as_mut().poll_next_request(cx)?) {
            Some(dispatch_request) => dispatch_request,
            None => return Poll::Ready(None),
        };
+        let entered = span.enter();
        // poll_next_request only returns Ready if there is room to buffer another request.
        // Therefore, we can call write_request without fear of erroring due to a full
        // buffer.
-        let request_id = dispatch_request.request_id;
+        let request_id = request_id;
        let request = ClientMessage::Request(Request {
            id: request_id,
-            message: dispatch_request.request,
+            message: request,
            context: context::Context {
-                deadline: dispatch_request.ctx.deadline,
-                trace_context: dispatch_request.ctx.trace_context,
+                deadline: ctx.deadline,
+                trace_context: ctx.trace_context,
            },
        });
-        self.transport_pin_mut().start_send(request)?;
+        self.start_send(request)?;
+        tracing::info!("SendRequest");
+        drop(entered);
+
        self.in_flight_requests()
-            .insert_request(
-                request_id,
-                dispatch_request.ctx,
-                dispatch_request.response_completion,
-            )
+            .insert_request(request_id, ctx, span, response_completion)
            .expect("Request IDs should be unique");
        Poll::Ready(Some(Ok(())))
    }

-    fn poll_write_cancel<'a>(self: &'a mut Pin<&mut Self>, cx: &mut Context<'_>) -> PollIo<()> {
-        let (context, request_id) = match ready!(self.as_mut().poll_next_cancellation(cx)?) {
-            Some((context, request_id)) => (context, request_id),
+    fn poll_write_cancel<'a>(
+        self: &'a mut Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Option<Result<(), ChannelError<C::Error>>>> {
+        let (context, span, request_id) = match ready!(self.as_mut().poll_next_cancellation(cx)?) {
+            Some(triple) => triple,
            None => return Poll::Ready(None),
        };
+        let _entered = span.enter();

-        let trace_id = *context.trace_id();
        let cancel = ClientMessage::Cancel {
            trace_context: context.trace_context,
            request_id,
        };
-        self.transport_pin_mut().start_send(cancel)?;
-        trace!("[{}] Cancel message sent.", trace_id);
+        self.start_send(cancel)?;
+        tracing::info!("CancelRequest");
        Poll::Ready(Some(Ok(())))
    }

@@ -460,28 +558,24 @@ impl<Req, Resp, C> Future for RequestDispatch<Req, Resp, C>
 where
    C: Transport<ClientMessage<Req>, Response<Resp>>,
 {
-    type Output = anyhow::Result<()>;
+    type Output = Result<(), ChannelError<C::Error>>;

-    fn poll(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<anyhow::Result<()>> {
+    fn poll(
+        mut self: Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Result<(), ChannelError<C::Error>>> {
        loop {
-            match (
-                self.as_mut()
-                    .pump_read(cx)
-                    .context("failed to read from transport")?,
-                self.as_mut()
-                    .pump_write(cx)
-                    .context("failed to write to transport")?,
-            ) {
+            match (self.as_mut().pump_read(cx)?, self.as_mut().pump_write(cx)?) {
                (Poll::Ready(None), _) => {
-                    info!("Shutdown: read half closed, so shutting down.");
+                    tracing::info!("Shutdown: read half closed, so shutting down.");
                    return Poll::Ready(Ok(()));
                }
                (read, Poll::Ready(None)) => {
-                    if self.in_flight_requests().is_empty() {
-                        info!("Shutdown: write half closed, and no requests in flight.");
+                    if self.in_flight_requests.is_empty() {
+                        tracing::info!("Shutdown: write half closed, and no requests in flight.");
                        return Poll::Ready(Ok(()));
                    }
-                    info!(
+                    tracing::info!(
                        "Shutdown: write half closed, and {} requests in flight.",
                        self.in_flight_requests().len()
                    );
@@ -502,114 +596,103 @@ where
 #[derive(Debug)]
 struct DispatchRequest<Req, Resp> {
    pub ctx: context::Context,
+    pub span: Span,
    pub request_id: u64,
    pub request: Req,
-    pub response_completion: oneshot::Sender<Response<Resp>>,
-}
-
-/// Sends request cancellation signals.
-#[derive(Debug, Clone)]
-struct RequestCancellation(mpsc::UnboundedSender<u64>);
-
-/// A stream of IDs of requests that have been canceled.
-#[derive(Debug)]
-struct CanceledRequests(mpsc::UnboundedReceiver<u64>);
-
-/// Returns a channel to send request cancellation messages.
-fn cancellations() -> (RequestCancellation, CanceledRequests) {
-    // Unbounded because messages are sent in the drop fn. This is fine, because it's still
-    // bounded by the number of in-flight requests. Additionally, each request has a clone
-    // of the sender, so the bounded channel would have the same behavior,
-    // since it guarantees a slot.
-    let (tx, rx) = mpsc::unbounded_channel();
-    (RequestCancellation(tx), CanceledRequests(rx))
-}
-
-impl RequestCancellation {
-    /// Cancels the request with ID `request_id`.
-    fn cancel(&mut self, request_id: u64) {
-        let _ = self.0.send(request_id);
-    }
-}
-
-impl CanceledRequests {
-    fn poll_recv(&mut self, cx: &mut Context<'_>) -> Poll<Option<u64>> {
-        self.0.poll_recv(cx)
-    }
-}
-
-impl Stream for CanceledRequests {
-    type Item = u64;
-
-    fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<u64>> {
-        self.poll_recv(cx)
-    }
+    pub response_completion: oneshot::Sender<Result<Response<Resp>, DeadlineExceededError>>,
 }

 #[cfg(test)]
 mod tests {
-    use super::{
-        cancellations, CanceledRequests, Channel, DispatchResponse, RequestCancellation,
-        RequestDispatch,
-    };
+    use super::{cancellations, Channel, DispatchRequest, RequestDispatch, ResponseGuard};
    use crate::{
-        client::{in_flight_requests::InFlightRequests, Config},
+        client::{
+            in_flight_requests::{DeadlineExceededError, InFlightRequests},
+            Config,
+        },
        context,
        transport::{self, channel::UnboundedChannel},
        ClientMessage, Response,
    };
+    use assert_matches::assert_matches;
    use futures::{prelude::*, task::*};
-    use std::{pin::Pin, sync::atomic::AtomicUsize, sync::Arc};
+    use std::{
+        convert::TryFrom,
+        pin::Pin,
+        sync::atomic::{AtomicUsize, Ordering},
+        sync::Arc,
+    };
    use tokio::sync::{mpsc, oneshot};
+    use tracing::Span;
+
+    #[tokio::test]
+    async fn response_completes_request_future() {
+        let (mut dispatch, mut _channel, mut server_channel) = set_up();
+        let cx = &mut Context::from_waker(&noop_waker_ref());
+        let (tx, mut rx) = oneshot::channel();
+
+        dispatch
+            .in_flight_requests
+            .insert_request(0, context::current(), Span::current(), tx)
+            .unwrap();
+        server_channel
+            .send(Response {
+                request_id: 0,
+                message: Ok("Resp".into()),
+            })
+            .await
+            .unwrap();
+        assert_matches!(dispatch.as_mut().poll(cx), Poll::Pending);
+        assert_matches!(rx.try_recv(), Ok(Ok(Response { request_id: 0, message: Ok(resp) })) if resp == "Resp");
+    }

    #[tokio::test]
    async fn dispatch_response_cancels_on_drop() {
        let (cancellation, mut canceled_requests) = cancellations();
-        let (_, response) = oneshot::channel();
-        drop(DispatchResponse::<u32> {
-            response,
-            cancellation: Some(cancellation),
+        let (_, mut response) = oneshot::channel();
+        drop(ResponseGuard::<u32> {
+            response: &mut response,
+            cancellation: &cancellation,
            request_id: 3,
-            ctx: context::current(),
        });
        // resp's drop() is run, which should send a cancel message.
        let cx = &mut Context::from_waker(&noop_waker_ref());
-        assert_eq!(canceled_requests.0.poll_recv(cx), Poll::Ready(Some(3)));
+        assert_eq!(canceled_requests.poll_recv(cx), Poll::Ready(Some(3)));
    }

    #[tokio::test]
    async fn dispatch_response_doesnt_cancel_after_complete() {
        let (cancellation, mut canceled_requests) = cancellations();
-        let (tx, response) = oneshot::channel();
-        tx.send(Response {
+        let (tx, mut response) = oneshot::channel();
+        tx.send(Ok(Response {
            request_id: 0,
            message: Ok("well done"),
-        })
+        }))
        .unwrap();
-        {
-            DispatchResponse {
-                response,
-                cancellation: Some(cancellation),
-                request_id: 3,
-                ctx: context::current(),
-            }
-            .await
-            .unwrap();
-            // resp's drop() is run, but should not send a cancel message.
+        // resp's drop() is run, but should not send a cancel message.
+        ResponseGuard {
+            response: &mut response,
+            cancellation: &cancellation,
+            request_id: 3,
        }
+        .response()
+        .await
+        .unwrap();
+        drop(cancellation);
        let cx = &mut Context::from_waker(&noop_waker_ref());
-        assert_eq!(canceled_requests.0.poll_recv(cx), Poll::Ready(None));
+        assert_eq!(canceled_requests.poll_recv(cx), Poll::Ready(None));
    }

    #[tokio::test]
    async fn stage_request() {
        let (mut dispatch, mut channel, _server_channel) = set_up();
-        let dispatch = Pin::new(&mut dispatch);
        let cx = &mut Context::from_waker(&noop_waker_ref());
+        let (tx, mut rx) = oneshot::channel();

-        let _resp = send_request(&mut channel, "hi").await;
+        let _resp = send_request(&mut channel, "hi", tx, &mut rx).await;

-        let req = dispatch.poll_next_request(cx).ready();
+        #[allow(unstable_name_collisions)]
+        let req = dispatch.as_mut().poll_next_request(cx).ready();
        assert!(req.is_some());

        let req = req.unwrap();
@@ -621,10 +704,10 @@ mod tests {
    #[tokio::test]
    async fn stage_request_channel_dropped_doesnt_panic() {
        let (mut dispatch, mut channel, mut server_channel) = set_up();
-        let mut dispatch = Pin::new(&mut dispatch);
        let cx = &mut Context::from_waker(&noop_waker_ref());
+        let (tx, mut rx) = oneshot::channel();

-        let _ = send_request(&mut channel, "hi").await;
+        let _ = send_request(&mut channel, "hi", tx, &mut rx).await;
        drop(channel);

        assert!(dispatch.as_mut().poll(cx).is_ready());
@@ -639,95 +722,117 @@ mod tests {
        dispatch.await.unwrap();
    }

+    #[allow(unstable_name_collisions)]
    #[tokio::test]
    async fn stage_request_response_future_dropped_is_canceled_before_sending() {
        let (mut dispatch, mut channel, _server_channel) = set_up();
-        let dispatch = Pin::new(&mut dispatch);
        let cx = &mut Context::from_waker(&noop_waker_ref());
+        let (tx, mut rx) = oneshot::channel();

-        let _ = send_request(&mut channel, "hi").await;
+        let _ = send_request(&mut channel, "hi", tx, &mut rx).await;

        // Drop the channel so polling returns none if no requests are currently ready.
        drop(channel);
        // Test that a request future dropped before it's processed by dispatch will cause the request
        // to not be added to the in-flight request map.
-        assert!(dispatch.poll_next_request(cx).ready().is_none());
+        assert!(dispatch.as_mut().poll_next_request(cx).ready().is_none());
    }

+    #[allow(unstable_name_collisions)]
    #[tokio::test]
    async fn stage_request_response_future_dropped_is_canceled_after_sending() {
        let (mut dispatch, mut channel, _server_channel) = set_up();
        let cx = &mut Context::from_waker(&noop_waker_ref());
-        let mut dispatch = Pin::new(&mut dispatch);
+        let (tx, mut rx) = oneshot::channel();

-        let req = send_request(&mut channel, "hi").await;
+        let req = send_request(&mut channel, "hi", tx, &mut rx).await;

        assert!(dispatch.as_mut().pump_write(cx).ready().is_some());
-        assert!(!dispatch.in_flight_requests().is_empty());
+        assert!(!dispatch.in_flight_requests.is_empty());

        // Test that a request future dropped after it's processed by dispatch will cause the request
        // to be removed from the in-flight request map.
        drop(req);
-        if let Poll::Ready(Some(_)) = dispatch.as_mut().poll_next_cancellation(cx).unwrap() {
-            // ok
-        } else {
-            panic!("Expected request to be cancelled")
-        };
-        assert!(dispatch.in_flight_requests().is_empty());
+        assert_matches!(
+            dispatch.as_mut().poll_next_cancellation(cx),
+            Poll::Ready(Some(Ok(_)))
+        );
+        assert!(dispatch.in_flight_requests.is_empty());
    }

    #[tokio::test]
    async fn stage_request_response_closed_skipped() {
        let (mut dispatch, mut channel, _server_channel) = set_up();
-        let dispatch = Pin::new(&mut dispatch);
        let cx = &mut Context::from_waker(&noop_waker_ref());
+        let (tx, mut rx) = oneshot::channel();

        // Test that a request future that's closed its receiver but not yet canceled its request --
        // i.e. still in `drop fn` -- will cause the request to not be added to the in-flight request
        // map.
-        let mut resp = send_request(&mut channel, "hi").await;
+        let resp = send_request(&mut channel, "hi", tx, &mut rx).await;
        resp.response.close();

-        assert!(dispatch.poll_next_request(cx).is_pending());
+        assert!(dispatch.as_mut().poll_next_request(cx).is_pending());
    }

    fn set_up() -> (
-        RequestDispatch<String, String, UnboundedChannel<Response<String>, ClientMessage<String>>>,
+        Pin<
+            Box<
+                RequestDispatch<
+                    String,
+                    String,
+                    UnboundedChannel<Response<String>, ClientMessage<String>>,
+                >,
+            >,
+        >,
        Channel<String, String>,
        UnboundedChannel<ClientMessage<String>, Response<String>>,
    ) {
-        let _ = env_logger::try_init();
+        let _ = tracing_subscriber::fmt().with_test_writer().try_init();

        let (to_dispatch, pending_requests) = mpsc::channel(1);
-        let (cancel_tx, canceled_requests) = mpsc::unbounded_channel();
+        let (cancellation, canceled_requests) = cancellations();
        let (client_channel, server_channel) = transport::channel::unbounded();

        let dispatch = RequestDispatch::<String, String, _> {
            transport: client_channel.fuse(),
            pending_requests: pending_requests,
-            canceled_requests: CanceledRequests(canceled_requests),
+            canceled_requests,
            in_flight_requests: InFlightRequests::default(),
            config: Config::default(),
        };

-        let cancellation = RequestCancellation(cancel_tx);
        let channel = Channel {
            to_dispatch,
            cancellation,
            next_request_id: Arc::new(AtomicUsize::new(0)),
        };

-        (dispatch, channel, server_channel)
+        (Box::pin(dispatch), channel, server_channel)
    }

-    async fn send_request(
-        channel: &mut Channel<String, String>,
+    async fn send_request<'a>(
+        channel: &'a mut Channel<String, String>,
        request: &str,
-    ) -> DispatchResponse<String> {
-        channel
-            .send(context::current(), request.to_string())
-            .await
-            .unwrap()
+        response_completion: oneshot::Sender<Result<Response<String>, DeadlineExceededError>>,
+        response: &'a mut oneshot::Receiver<Result<Response<String>, DeadlineExceededError>>,
+    ) -> ResponseGuard<'a, String> {
+        let request_id =
+            u64::try_from(channel.next_request_id.fetch_add(1, Ordering::Relaxed)).unwrap();
+        let request = DispatchRequest {
+            ctx: context::current(),
+            span: Span::current(),
+            request_id,
+            request: request.to_string(),
+            response_completion,
+        };
+        let response_guard = ResponseGuard {
+            response,
+            cancellation: &channel.cancellation,
+            request_id,
+        };
+        channel.to_dispatch.send(request).await.unwrap();
+        response_guard
    }

    async fn send_response(
--- a/tarpc/src/client/in_flight_requests.rs
+++ b/tarpc/src/client/in_flight_requests.rs
@@ -1,18 +1,16 @@
 use crate::{
    context,
    util::{Compact, TimeUntil},
-    PollIo, Response, ServerError,
+    Response,
 };
 use fnv::FnvHashMap;
-use futures::ready;
-use log::{debug, trace};
 use std::{
    collections::hash_map,
-    io,
    task::{Context, Poll},
 };
 use tokio::sync::oneshot;
 use tokio_util::time::delay_queue::{self, DelayQueue};
+use tracing::Span;

 /// Requests already written to the wire that haven't yet received responses.
 #[derive(Debug)]
@@ -30,10 +28,17 @@ impl<Resp> Default for InFlightRequests<Resp> {
    }
 }

+/// The request exceeded its deadline.
+#[derive(thiserror::Error, Debug)]
+#[non_exhaustive]
+#[error("the request exceeded its deadline")]
+pub struct DeadlineExceededError;
+
 #[derive(Debug)]
 struct RequestData<Resp> {
    ctx: context::Context,
-    response_completion: oneshot::Sender<Response<Resp>>,
+    span: Span,
+    response_completion: oneshot::Sender<Result<Response<Resp>, DeadlineExceededError>>,
    /// The key to remove the timer for the request's deadline.
    deadline_key: delay_queue::Key,
 }
@@ -59,20 +64,16 @@ impl<Resp> InFlightRequests<Resp> {
        &mut self,
        request_id: u64,
        ctx: context::Context,
-        response_completion: oneshot::Sender<Response<Resp>>,
+        span: Span,
+        response_completion: oneshot::Sender<Result<Response<Resp>, DeadlineExceededError>>,
    ) -> Result<(), AlreadyExistsError> {
        match self.request_data.entry(request_id) {
            hash_map::Entry::Vacant(vacant) => {
                let timeout = ctx.deadline.time_until();
-                trace!(
-                    "[{}] Queuing request with timeout {:?}.",
-                    ctx.trace_id(),
-                    timeout,
-                );
-
                let deadline_key = self.deadlines.insert(request_id, timeout);
                vacant.insert(RequestData {
                    ctx,
+                    span,
                    response_completion,
                    deadline_key,
                });
@@ -85,15 +86,15 @@ impl<Resp> InFlightRequests<Resp> {
    /// Removes a request without aborting. Returns true iff the request was found.
    pub fn complete_request(&mut self, response: Response<Resp>) -> bool {
        if let Some(request_data) = self.request_data.remove(&response.request_id) {
+            let _entered = request_data.span.enter();
+            tracing::info!("ReceiveResponse");
            self.request_data.compact(0.1);
-
-            trace!("[{}] Received response.", request_data.ctx.trace_id());
            self.deadlines.remove(&request_data.deadline_key);
-            request_data.complete(response);
+            let _ = request_data.response_completion.send(Ok(response));
            return true;
        }

-        debug!(
+        tracing::debug!(
            "No in-flight request found for request_id = {}.",
            response.request_id
        );
@@ -104,12 +105,11 @@ impl<Resp> InFlightRequests<Resp> {

    /// Cancels a request without completing (typically used when a request handle was dropped
    /// before the request completed).
-    pub fn cancel_request(&mut self, request_id: u64) -> Option<context::Context> {
+    pub fn cancel_request(&mut self, request_id: u64) -> Option<(context::Context, Span)> {
        if let Some(request_data) = self.request_data.remove(&request_id) {
            self.request_data.compact(0.1);
-            trace!("[{}] Cancelling request.", request_data.ctx.trace_id());
            self.deadlines.remove(&request_data.deadline_key);
-            Some(request_data.ctx)
+            Some((request_data.ctx, request_data.span))
        } else {
            None
        }
@@ -117,46 +117,18 @@ impl<Resp> InFlightRequests<Resp> {

    /// Yields a request that has expired, completing it with a TimedOut error.
    /// The caller should send cancellation messages for any yielded request ID.
-    pub fn poll_expired(&mut self, cx: &mut Context) -> PollIo<u64> {
-        Poll::Ready(match ready!(self.deadlines.poll_expired(cx)) {
-            Some(Ok(expired)) => {
-                let request_id = expired.into_inner();
-                if let Some(request_data) = self.request_data.remove(&request_id) {
-                    self.request_data.compact(0.1);
-                    request_data.complete(Self::deadline_exceeded_error(request_id));
-                }
-                Some(Ok(request_id))
+    pub fn poll_expired(&mut self, cx: &mut Context) -> Poll<Option<u64>> {
+        self.deadlines.poll_expired(cx).map(|expired| {
+            let request_id = expired?.into_inner();
+            if let Some(request_data) = self.request_data.remove(&request_id) {
+                let _entered = request_data.span.enter();
+                tracing::error!("DeadlineExceeded");
+                self.request_data.compact(0.1);
+                let _ = request_data
+                    .response_completion
+                    .send(Err(DeadlineExceededError));
            }
-            Some(Err(e)) => Some(Err(io::Error::new(io::ErrorKind::Other, e))),
-            None => None,
+            Some(request_id)
        })
    }
-
-    fn deadline_exceeded_error(request_id: u64) -> Response<Resp> {
-        Response {
-            request_id,
-            message: Err(ServerError {
-                kind: io::ErrorKind::TimedOut,
-                detail: Some("Client dropped expired request.".to_string()),
-            }),
-        }
-    }
-}
-
-/// When InFlightRequests is dropped, any outstanding requests are completed with a
-/// deadline-exceeded error.
-impl<Resp> Drop for InFlightRequests<Resp> {
-    fn drop(&mut self) {
-        let deadlines = &mut self.deadlines;
-        for (_, request_data) in self.request_data.drain() {
-            let expired = deadlines.remove(&request_data.deadline_key);
-            request_data.complete(Self::deadline_exceeded_error(expired.into_inner()));
-        }
-    }
-}
-
-impl<Resp> RequestData<Resp> {
-    fn complete(self, response: Response<Resp>) {
-        let _ = self.response_completion.send(response);
-    }
 }
--- a/tarpc/src/context.rs
+++ b/tarpc/src/context.rs
@@ -8,8 +8,13 @@
 //! client to server and is used by the server to enforce response deadlines.

 use crate::trace::{self, TraceId};
+use opentelemetry::trace::TraceContextExt;
 use static_assertions::assert_impl_all;
-use std::time::{Duration, SystemTime};
+use std::{
+    convert::TryFrom,
+    time::{Duration, SystemTime},
+};
+use tracing_opentelemetry::OpenTelemetrySpanExt;

 /// A request context that carries request-scoped information like deadlines and trace information.
 /// It is sent from client to server and is used by the server to enforce response deadlines.
@@ -22,15 +27,9 @@ use std::time::{Duration, SystemTime};
 pub struct Context {
    /// When the client expects the request to be complete by. The server should cancel the request
    /// if it is not complete by this time.
-    #[cfg_attr(
-        feature = "serde1",
-        serde(serialize_with = "crate::util::serde::serialize_epoch_secs")
-    )]
-    #[cfg_attr(
-        feature = "serde1",
-        serde(deserialize_with = "crate::util::serde::deserialize_epoch_secs")
-    )]
    #[cfg_attr(feature = "serde1", serde(default = "ten_seconds_from_now"))]
+    // Serialized as a Duration to prevent clock skew issues.
+    #[cfg_attr(feature = "serde1", serde(with = "absolute_to_relative_time"))]
    pub deadline: SystemTime,
    /// Uniquely identifies requests originating from the same source.
    /// When a service handles a request by making requests itself, those requests should
@@ -39,25 +38,115 @@ pub struct Context {
    pub trace_context: trace::Context,
 }

+#[cfg(feature = "serde1")]
+mod absolute_to_relative_time {
+    pub use serde::{Deserialize, Deserializer, Serialize, Serializer};
+    pub use std::time::{Duration, SystemTime};
+
+    pub fn serialize<S>(deadline: &SystemTime, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: Serializer,
+    {
+        let deadline = deadline
+            .duration_since(SystemTime::now())
+            .unwrap_or(Duration::ZERO);
+        deadline.serialize(serializer)
+    }
+
+    pub fn deserialize<'de, D>(deserializer: D) -> Result<SystemTime, D::Error>
+    where
+        D: Deserializer<'de>,
+    {
+        let deadline = Duration::deserialize(deserializer)?;
+        Ok(SystemTime::now() + deadline)
+    }
+
+    #[cfg(test)]
+    #[derive(serde::Serialize, serde::Deserialize)]
+    struct AbsoluteToRelative(#[serde(with = "self")] SystemTime);
+
+    #[test]
+    fn test_serialize() {
+        let now = SystemTime::now();
+        let deadline = now + Duration::from_secs(10);
+        let serialized_deadline = bincode::serialize(&AbsoluteToRelative(deadline)).unwrap();
+        let deserialized_deadline: Duration = bincode::deserialize(&serialized_deadline).unwrap();
+        // TODO: how to avoid flakiness?
+        assert!(deserialized_deadline > Duration::from_secs(9));
+    }
+
+    #[test]
+    fn test_deserialize() {
+        let deadline = Duration::from_secs(10);
+        let serialized_deadline = bincode::serialize(&deadline).unwrap();
+        let AbsoluteToRelative(deserialized_deadline) =
+            bincode::deserialize(&serialized_deadline).unwrap();
+        // TODO: how to avoid flakiness?
+        assert!(deserialized_deadline > SystemTime::now() + Duration::from_secs(9));
+    }
+}
+
 assert_impl_all!(Context: Send, Sync);

-#[cfg(feature = "serde1")]
 fn ten_seconds_from_now() -> SystemTime {
    SystemTime::now() + Duration::from_secs(10)
 }

 /// Returns the context for the current request, or a default Context if no request is active.
-// TODO: populate Context with request-scoped data, with default fallbacks.
 pub fn current() -> Context {
-    Context {
-        deadline: SystemTime::now() + Duration::from_secs(10),
-        trace_context: trace::Context::new_root(),
+    Context::current()
+}
+
+#[derive(Clone)]
+struct Deadline(SystemTime);
+
+impl Default for Deadline {
+    fn default() -> Self {
+        Self(ten_seconds_from_now())
    }
 }

 impl Context {
+    /// Returns the context for the current request, or a default Context if no request is active.
+    pub fn current() -> Self {
+        let span = tracing::Span::current();
+        Self {
+            trace_context: trace::Context::try_from(&span)
+                .unwrap_or_else(|_| trace::Context::default()),
+            deadline: span
+                .context()
+                .get::<Deadline>()
+                .cloned()
+                .unwrap_or_default()
+                .0,
+        }
+    }
+
    /// Returns the ID of the request-scoped trace.
    pub fn trace_id(&self) -> &TraceId {
        &self.trace_context.trace_id
    }
 }
+
+/// An extension trait for [`tracing::Span`] for propagating tarpc Contexts.
+pub(crate) trait SpanExt {
+    /// Sets the given context on this span. Newly-created spans will be children of the given
+    /// context's trace context.
+    fn set_context(&self, context: &Context);
+}
+
+impl SpanExt for tracing::Span {
+    fn set_context(&self, context: &Context) {
+        self.set_parent(
+            opentelemetry::Context::new()
+                .with_remote_span_context(opentelemetry::trace::SpanContext::new(
+                    opentelemetry::trace::TraceId::from(context.trace_context.trace_id),
+                    opentelemetry::trace::SpanId::from(context.trace_context.span_id),
+                    opentelemetry::trace::TraceFlags::from(context.trace_context.sampling_decision),
+                    true,
+                    opentelemetry::trace::TraceState::default(),
+                ))
+                .with_value(Deadline(context.deadline)),
+        );
+    }
+}
--- a/tarpc/src/lib.rs
+++ b/tarpc/src/lib.rs
@@ -27,7 +27,7 @@
 //! process, and no context switching between different languages.
 //!
 //! Some other features of tarpc:
-//! - Pluggable transport: any type impling `Stream<Item = Request> + Sink<Response>` can be
+//! - Pluggable transport: any type implementing `Stream<Item = Request> + Sink<Response>` can be
 //!   used as a transport to connect the client and server.
 //! - `Send + 'static` optional: if the transport doesn't require it, neither does tarpc!
 //! - Cascading cancellation: dropping a request will send a cancellation message to the server.
@@ -38,6 +38,14 @@
 //!   requests sent by the server that use the request context will propagate the request deadline.
 //!   For example, if a server is handling a request with a 10s deadline, does 2s of work, then
 //!   sends a request to another server, that server will see an 8s deadline.
+//! - Distributed tracing: tarpc is instrumented with
+//!   [tracing](https://github.com/tokio-rs/tracing) primitives extended with
+//!   [OpenTelemetry](https://opentelemetry.io/) traces. Using a compatible tracing subscriber like
+//!   [Jaeger](https://github.com/open-telemetry/opentelemetry-rust/tree/main/opentelemetry-jaeger),
+//!   each RPC can be traced through the client, server, and other dependencies downstream of the
+//!   server. Even for applications not connected to a distributed tracing collector, the
+//!   instrumentation can also be ingested by regular loggers like
+//!   [env_logger](https://github.com/env-logger-rs/env_logger/).
 //! - Serde serialization: enabling the `serde1` Cargo feature will make service requests and
 //!   responses `Serialize + Deserialize`. It's entirely optional, though: in-memory transports can
 //!   be used, as well, so the price of serialization doesn't have to be paid when it's not needed.
@@ -46,7 +54,7 @@
 //! Add to your `Cargo.toml` dependencies:
 //!
 //! ```toml
-//! tarpc = "0.25"
+//! tarpc = "0.29"
 //! ```
 //!
 //! The `tarpc::service` attribute expands to a collection of items that form an rpc service.
@@ -59,8 +67,9 @@
 //! your `Cargo.toml`:
 //!
 //! ```toml
-//! futures = "1.0"
-//! tarpc = { version = "0.25", features = ["tokio1"] }
+//! anyhow = "1.0"
+//! futures = "0.3"
+//! tarpc = { version = "0.29", features = ["tokio1"] }
 //! tokio = { version = "1.0", features = ["macros"] }
 //! ```
 //!
@@ -79,9 +88,8 @@
 //! };
 //! use tarpc::{
 //!     client, context,
-//!     server::{self, Incoming},
+//!     server::{self, incoming::Incoming, Channel},
 //! };
-//! use std::io;
 //!
 //! // This is the service definition. It looks a lot like a trait definition.
 //! // It defines one RPC, hello, which takes one arg, name, and returns a String.
@@ -103,9 +111,8 @@
 //! # };
 //! # use tarpc::{
 //! #     client, context,
-//! #     server::{self, Incoming},
+//! #     server::{self, incoming::Incoming},
 //! # };
-//! # use std::io;
 //! # // This is the service definition. It looks a lot like a trait definition.
 //! # // It defines one RPC, hello, which takes one arg, name, and returns a String.
 //! # #[tarpc::service]
@@ -125,7 +132,7 @@
 //!     type HelloFut = Ready<String>;
 //!
 //!     fn hello(self, _: context::Context, name: String) -> Self::HelloFut {
-//!         future::ready(format!("Hello, {}!", name))
+//!         future::ready(format!("Hello, {name}!"))
 //!     }
 //! }
 //! ```
@@ -145,7 +152,6 @@
 //! #     client, context,
 //! #     server::{self, Channel},
 //! # };
-//! # use std::io;
 //! # // This is the service definition. It looks a lot like a trait definition.
 //! # // It defines one RPC, hello, which takes one arg, name, and returns a String.
 //! # #[tarpc::service]
@@ -162,14 +168,14 @@
 //! #     // an associated type representing the future output by the fn.
 //! #     type HelloFut = Ready<String>;
 //! #     fn hello(self, _: context::Context, name: String) -> Self::HelloFut {
-//! #         future::ready(format!("Hello, {}!", name))
+//! #         future::ready(format!("Hello, {name}!"))
 //! #     }
 //! # }
 //! # #[cfg(not(feature = "tokio1"))]
 //! # fn main() {}
 //! # #[cfg(feature = "tokio1")]
 //! #[tokio::main]
-//! async fn main() -> io::Result<()> {
+//! async fn main() -> anyhow::Result<()> {
 //!     let (client_transport, server_transport) = tarpc::transport::channel::unbounded();
 //!
 //!     let server = server::BaseChannel::with_defaults(server_transport);
@@ -177,14 +183,14 @@
 //!
 //!     // WorldClient is generated by the #[tarpc::service] attribute. It has a constructor `new`
 //!     // that takes a config and any Transport as input.
-//!     let mut client = WorldClient::new(client::Config::default(), client_transport).spawn()?;
+//!     let mut client = WorldClient::new(client::Config::default(), client_transport).spawn();
 //!
 //!     // The client has an RPC method for each RPC defined in the annotated trait. It takes the same
 //!     // args as defined, with the addition of a Context, which is always the first arg. The Context
 //!     // specifies a deadline and trace information which can be helpful in debugging requests.
 //!     let hello = client.hello(context::current(), "Stim".to_string()).await?;
 //!
-//!     println!("{}", hello);
+//!     println!("{hello}");
 //!
 //!     Ok(())
 //! }
@@ -199,10 +205,11 @@
 #![cfg_attr(docsrs, feature(doc_cfg))]

 #[cfg(feature = "serde1")]
+#[doc(hidden)]
 pub use serde;

 #[cfg(feature = "serde-transport")]
-pub use tokio_serde;
+pub use {tokio_serde, tokio_util};

 #[cfg(feature = "serde-transport")]
 #[cfg_attr(docsrs, doc(cfg(feature = "serde-transport")))]
@@ -257,7 +264,7 @@ pub use tarpc_plugins::service;
 /// #[tarpc::server]
 /// impl World for HelloServer {
 ///     async fn hello(self, _: context::Context, name: String) -> String {
-///         format!("Hello, {}! You are connected from {:?}.", name, self.0)
+///         format!("Hello, {name}! You are connected from {:?}.", self.0)
 ///     }
 /// }
 /// ```
@@ -283,7 +290,7 @@ pub use tarpc_plugins::service;
 ///     fn hello(self, _: context::Context, name: String) -> Pin<Box<dyn Future<Output = String>
 ///     + Send>> {
 ///         Box::pin(async move {
-///             format!("Hello, {}! You are connected from {:?}.", name, self.0)
+///             format!("Hello, {name}! You are connected from {:?}.", self.0)
 ///         })
 ///     }
 /// }
@@ -293,6 +300,7 @@ pub use tarpc_plugins::service;
 /// `async`, meaning that this should not break existing code.
 pub use tarpc_plugins::server;

+pub(crate) mod cancellations;
 pub mod client;
 pub mod context;
 pub mod server;
@@ -303,7 +311,7 @@ pub use crate::transport::sealed::Transport;

 use anyhow::Context as _;
 use futures::task::*;
-use std::{fmt::Display, io, time::SystemTime};
+use std::{error::Error, fmt::Display, io, time::SystemTime};

 /// A message from a client to a server.
 #[derive(Debug)]
@@ -355,8 +363,9 @@ pub struct Response<T> {
    pub message: Result<T, ServerError>,
 }

-/// An error response from a server to a client.
-#[derive(Clone, Debug, PartialEq, Eq, Hash)]
+/// An error indicating the server aborted the request early, e.g., due to request throttling.
+#[derive(thiserror::Error, Clone, Debug, PartialEq, Eq, Hash)]
+#[error("{kind:?}: {detail}")]
 #[non_exhaustive]
 #[cfg_attr(feature = "serde1", derive(serde::Serialize, serde::Deserialize))]
 pub struct ServerError {
@@ -371,13 +380,7 @@ pub struct ServerError {
    /// The type of error that occurred to fail the request.
    pub kind: io::ErrorKind,
    /// A message describing more detail about the error that occurred.
-    pub detail: Option<String>,
-}
-
-impl From<ServerError> for io::Error {
-    fn from(e: ServerError) -> io::Error {
-        io::Error::new(e.kind, e.detail.unwrap_or_default())
-    }
+    pub detail: String,
 }

 impl<T> Request<T> {
@@ -387,7 +390,6 @@ impl<T> Request<T> {
    }
 }

-pub(crate) type PollIo<T> = Poll<Option<io::Result<T>>>;
 pub(crate) trait PollContext<T> {
    fn context<C>(self, context: C) -> Poll<Option<anyhow::Result<T>>>
    where
@@ -399,7 +401,10 @@ pub(crate) trait PollContext<T> {
        F: FnOnce() -> C;
 }

-impl<T> PollContext<T> for PollIo<T> {
+impl<T, E> PollContext<T> for Poll<Option<Result<T, E>>>
+where
+    E: Error + Send + Sync + 'static,
+{
    fn context<C>(self, context: C) -> Poll<Option<anyhow::Result<T>>>
    where
        C: Display + Send + Sync + 'static,
--- a/tarpc/src/serde_transport.rs
+++ b/tarpc/src/serde_transport.rs
@@ -42,14 +42,10 @@ where
    type Item = io::Result<Item>;

    fn poll_next(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Option<io::Result<Item>>> {
-        match self.project().inner.poll_next(cx) {
-            Poll::Pending => Poll::Pending,
-            Poll::Ready(None) => Poll::Ready(None),
-            Poll::Ready(Some(Ok::<_, CodecError>(next))) => Poll::Ready(Some(Ok(next))),
-            Poll::Ready(Some(Err::<_, CodecError>(e))) => {
-                Poll::Ready(Some(Err(io::Error::new(io::ErrorKind::Other, e))))
-            }
-        }
+        self.project()
+            .inner
+            .poll_next(cx)
+            .map_err(|e| io::Error::new(io::ErrorKind::Other, e))
    }
 }

@@ -65,7 +61,10 @@ where
    type Error = io::Error;

    fn poll_ready(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
-        convert(self.project().inner.poll_ready(cx))
+        self.project()
+            .inner
+            .poll_ready(cx)
+            .map_err(|e| io::Error::new(io::ErrorKind::Other, e))
    }

    fn start_send(self: Pin<&mut Self>, item: SinkItem) -> io::Result<()> {
@@ -76,20 +75,20 @@ where
    }

    fn poll_flush(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
-        convert(self.project().inner.poll_flush(cx))
+        self.project()
+            .inner
+            .poll_flush(cx)
+            .map_err(|e| io::Error::new(io::ErrorKind::Other, e))
    }

    fn poll_close(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
-        convert(self.project().inner.poll_close(cx))
+        self.project()
+            .inner
+            .poll_close(cx)
+            .map_err(|e| io::Error::new(io::ErrorKind::Other, e))
    }
 }

-fn convert<E: Into<Box<dyn Error + Send + Sync>>>(
-    poll: Poll<Result<(), E>>,
-) -> Poll<io::Result<()>> {
-    poll.map(|ready| ready.map_err(|e| io::Error::new(io::ErrorKind::Other, e)))
-}
-
 /// Constructs a new transport from a framed transport and a serialization codec.
 pub fn new<S, Item, SinkItem, Codec>(
    framed_io: Framed<S, LengthDelimitedCodec>,
@@ -150,6 +149,7 @@ pub mod tcp {
    }

    /// A connection Future that also exposes the length-delimited framing config.
+    #[must_use]
    #[pin_project]
    pub struct Connect<T, Item, SinkItem, CodecFn> {
        #[pin]
--- a/tarpc/src/server.rs
+++ b/tarpc/src/server.rs
--- a/tarpc/src/server/in_flight_requests.rs
+++ b/tarpc/src/server/in_flight_requests.rs
@@ -1,19 +1,13 @@
-use crate::{
-    util::{Compact, TimeUntil},
-    PollIo,
-};
+use crate::util::{Compact, TimeUntil};
 use fnv::FnvHashMap;
-use futures::{
-    future::{AbortHandle, AbortRegistration},
-    ready,
-};
+use futures::future::{AbortHandle, AbortRegistration};
 use std::{
    collections::hash_map,
-    io,
    task::{Context, Poll},
    time::SystemTime,
 };
 use tokio_util::time::delay_queue::{self, DelayQueue};
+use tracing::Span;

 /// A data structure that tracks in-flight requests. It aborts requests,
 /// either on demand or when a request deadline expires.
@@ -23,13 +17,15 @@ pub struct InFlightRequests {
    deadlines: DelayQueue<u64>,
 }

-#[derive(Debug)]
 /// Data needed to clean up a single in-flight request.
+#[derive(Debug)]
 struct RequestData {
    /// Aborts the response handler for the associated request.
    abort_handle: AbortHandle,
    /// The key to remove the timer for the request's deadline.
    deadline_key: delay_queue::Key,
+    /// The client span.
+    span: Span,
 }

 /// An error returned when a request attempted to start with the same ID as a request already
@@ -48,6 +44,7 @@ impl InFlightRequests {
        &mut self,
        request_id: u64,
        deadline: SystemTime,
+        span: Span,
    ) -> Result<AbortRegistration, AlreadyExistsError> {
        match self.request_data.entry(request_id) {
            hash_map::Entry::Vacant(vacant) => {
@@ -57,6 +54,7 @@ impl InFlightRequests {
                vacant.insert(RequestData {
                    abort_handle,
                    deadline_key,
+                    span,
                });
                Ok(abort_registration)
            }
@@ -66,12 +64,17 @@ impl InFlightRequests {

    /// Cancels an in-flight request. Returns true iff the request was found.
    pub fn cancel_request(&mut self, request_id: u64) -> bool {
-        if let Some(request_data) = self.request_data.remove(&request_id) {
+        if let Some(RequestData {
+            span,
+            abort_handle,
+            deadline_key,
+        }) = self.request_data.remove(&request_id)
+        {
+            let _entered = span.enter();
            self.request_data.compact(0.1);
-
-            request_data.abort_handle.abort();
-            self.deadlines.remove(&request_data.deadline_key);
-
+            abort_handle.abort();
+            self.deadlines.remove(&deadline_key);
+            tracing::info!("ReceiveCancel");
            true
        } else {
            false
@@ -80,30 +83,35 @@ impl InFlightRequests {

    /// Removes a request without aborting. Returns true iff the request was found.
    /// This method should be used when a response is being sent.
-    pub fn remove_request(&mut self, request_id: u64) -> bool {
+    pub fn remove_request(&mut self, request_id: u64) -> Option<Span> {
        if let Some(request_data) = self.request_data.remove(&request_id) {
            self.request_data.compact(0.1);
-
            self.deadlines.remove(&request_data.deadline_key);
-
-            true
+            Some(request_data.span)
        } else {
-            false
+            None
        }
    }

    /// Yields a request that has expired, aborting any ongoing processing of that request.
-    pub fn poll_expired(&mut self, cx: &mut Context) -> PollIo<u64> {
-        Poll::Ready(match ready!(self.deadlines.poll_expired(cx)) {
-            Some(Ok(expired)) => {
-                if let Some(request_data) = self.request_data.remove(expired.get_ref()) {
-                    self.request_data.compact(0.1);
-                    request_data.abort_handle.abort();
-                }
-                Some(Ok(expired.into_inner()))
+    pub fn poll_expired(&mut self, cx: &mut Context) -> Poll<Option<u64>> {
+        if self.deadlines.is_empty() {
+            // TODO(https://github.com/tokio-rs/tokio/issues/4161)
+            // This is a workaround for DelayQueue not always treating this case correctly.
+            return Poll::Ready(None);
+        }
+        self.deadlines.poll_expired(cx).map(|expired| {
+            let expired = expired?;
+            if let Some(RequestData {
+                abort_handle, span, ..
+            }) = self.request_data.remove(expired.get_ref())
+            {
+                let _entered = span.enter();
+                self.request_data.compact(0.1);
+                abort_handle.abort();
+                tracing::error!("DeadlineExceeded");
            }
-            Some(Err(e)) => Some(Err(io::Error::new(io::ErrorKind::Other, e))),
-            None => None,
+            Some(expired.into_inner())
        })
    }
 }
@@ -118,75 +126,96 @@ impl Drop for InFlightRequests {
 }

 #[cfg(test)]
-use {
-    assert_matches::assert_matches,
-    futures::{
+mod tests {
+    use super::*;
+
+    use assert_matches::assert_matches;
+    use futures::{
        future::{pending, Abortable},
        FutureExt,
-    },
-    futures_test::task::noop_context,
-};
+    };
+    use futures_test::task::noop_context;

-#[tokio::test]
-async fn start_request_increases_len() {
-    let mut in_flight_requests = InFlightRequests::default();
-    assert_eq!(in_flight_requests.len(), 0);
-    in_flight_requests
-        .start_request(0, SystemTime::now())
-        .unwrap();
-    assert_eq!(in_flight_requests.len(), 1);
-}
-
-#[tokio::test]
-async fn polling_expired_aborts() {
-    let mut in_flight_requests = InFlightRequests::default();
-    let abort_registration = in_flight_requests
-        .start_request(0, SystemTime::now())
-        .unwrap();
-    let mut abortable_future = Box::new(Abortable::new(pending::<()>(), abort_registration));
-
-    tokio::time::pause();
-    tokio::time::advance(std::time::Duration::from_secs(1000)).await;
-
-    assert_matches!(
-        in_flight_requests.poll_expired(&mut noop_context()),
-        Poll::Ready(Some(Ok(_)))
-    );
-    assert_matches!(
-        abortable_future.poll_unpin(&mut noop_context()),
-        Poll::Ready(Err(_))
-    );
-    assert_eq!(in_flight_requests.len(), 0);
-}
-
-#[tokio::test]
-async fn cancel_request_aborts() {
-    let mut in_flight_requests = InFlightRequests::default();
-    let abort_registration = in_flight_requests
-        .start_request(0, SystemTime::now())
-        .unwrap();
-    let mut abortable_future = Box::new(Abortable::new(pending::<()>(), abort_registration));
-
-    assert_eq!(in_flight_requests.cancel_request(0), true);
-    assert_matches!(
-        abortable_future.poll_unpin(&mut noop_context()),
-        Poll::Ready(Err(_))
-    );
-    assert_eq!(in_flight_requests.len(), 0);
-}
-
-#[tokio::test]
-async fn remove_request_doesnt_abort() {
-    let mut in_flight_requests = InFlightRequests::default();
-    let abort_registration = in_flight_requests
-        .start_request(0, SystemTime::now())
-        .unwrap();
-    let mut abortable_future = Box::new(Abortable::new(pending::<()>(), abort_registration));
-
-    assert_eq!(in_flight_requests.remove_request(0), true);
-    assert_matches!(
-        abortable_future.poll_unpin(&mut noop_context()),
-        Poll::Pending
-    );
-    assert_eq!(in_flight_requests.len(), 0);
+    #[tokio::test]
+    async fn start_request_increases_len() {
+        let mut in_flight_requests = InFlightRequests::default();
+        assert_eq!(in_flight_requests.len(), 0);
+        in_flight_requests
+            .start_request(0, SystemTime::now(), Span::current())
+            .unwrap();
+        assert_eq!(in_flight_requests.len(), 1);
+    }
+
+    #[tokio::test]
+    async fn polling_expired_aborts() {
+        let mut in_flight_requests = InFlightRequests::default();
+        let abort_registration = in_flight_requests
+            .start_request(0, SystemTime::now(), Span::current())
+            .unwrap();
+        let mut abortable_future = Box::new(Abortable::new(pending::<()>(), abort_registration));
+
+        tokio::time::pause();
+        tokio::time::advance(std::time::Duration::from_secs(1000)).await;
+
+        assert_matches!(
+            in_flight_requests.poll_expired(&mut noop_context()),
+            Poll::Ready(Some(_))
+        );
+        assert_matches!(
+            abortable_future.poll_unpin(&mut noop_context()),
+            Poll::Ready(Err(_))
+        );
+        assert_eq!(in_flight_requests.len(), 0);
+    }
+
+    #[tokio::test]
+    async fn cancel_request_aborts() {
+        let mut in_flight_requests = InFlightRequests::default();
+        let abort_registration = in_flight_requests
+            .start_request(0, SystemTime::now(), Span::current())
+            .unwrap();
+        let mut abortable_future = Box::new(Abortable::new(pending::<()>(), abort_registration));
+
+        assert_eq!(in_flight_requests.cancel_request(0), true);
+        assert_matches!(
+            abortable_future.poll_unpin(&mut noop_context()),
+            Poll::Ready(Err(_))
+        );
+        assert_eq!(in_flight_requests.len(), 0);
+    }
+
+    #[tokio::test]
+    async fn remove_request_doesnt_abort() {
+        let mut in_flight_requests = InFlightRequests::default();
+        assert!(in_flight_requests.deadlines.is_empty());
+
+        let abort_registration = in_flight_requests
+            .start_request(
+                0,
+                SystemTime::now() + std::time::Duration::from_secs(10),
+                Span::current(),
+            )
+            .unwrap();
+        let mut abortable_future = Box::new(Abortable::new(pending::<()>(), abort_registration));
+
+        // Precondition: Pending expiration
+        assert_matches!(
+            in_flight_requests.poll_expired(&mut noop_context()),
+            Poll::Pending
+        );
+        assert!(!in_flight_requests.deadlines.is_empty());
+
+        assert_matches!(in_flight_requests.remove_request(0), Some(_));
+        // Postcondition: No pending expirations
+        assert!(in_flight_requests.deadlines.is_empty());
+        assert_matches!(
+            in_flight_requests.poll_expired(&mut noop_context()),
+            Poll::Ready(None)
+        );
+        assert_matches!(
+            abortable_future.poll_unpin(&mut noop_context()),
+            Poll::Pending
+        );
+        assert_eq!(in_flight_requests.len(), 0);
+    }
 }
--- a/tarpc/src/server/incoming.rs
+++ b/tarpc/src/server/incoming.rs
@@ -0,0 +1,49 @@
+use super::{
+    limits::{channels_per_key::MaxChannelsPerKey, requests_per_channel::MaxRequestsPerChannel},
+    Channel,
+};
+use futures::prelude::*;
+use std::{fmt, hash::Hash};
+
+#[cfg(feature = "tokio1")]
+use super::{tokio::TokioServerExecutor, Serve};
+
+/// An extension trait for [streams](futures::prelude::Stream) of [`Channels`](Channel).
+pub trait Incoming<C>
+where
+    Self: Sized + Stream<Item = C>,
+    C: Channel,
+{
+    /// Enforces channel per-key limits.
+    fn max_channels_per_key<K, KF>(self, n: u32, keymaker: KF) -> MaxChannelsPerKey<Self, K, KF>
+    where
+        K: fmt::Display + Eq + Hash + Clone + Unpin,
+        KF: Fn(&C) -> K,
+    {
+        MaxChannelsPerKey::new(self, n, keymaker)
+    }
+
+    /// Caps the number of concurrent requests per channel.
+    fn max_concurrent_requests_per_channel(self, n: usize) -> MaxRequestsPerChannel<Self> {
+        MaxRequestsPerChannel::new(self, n)
+    }
+
+    /// [Executes](Channel::execute) each incoming channel. Each channel will be handled
+    /// concurrently by spawning on tokio's default executor, and each request will be also
+    /// be spawned on tokio's default executor.
+    #[cfg(feature = "tokio1")]
+    #[cfg_attr(docsrs, doc(cfg(feature = "tokio1")))]
+    fn execute<S>(self, serve: S) -> TokioServerExecutor<Self, S>
+    where
+        S: Serve<C::Req, Resp = C::Resp>,
+    {
+        TokioServerExecutor::new(self, serve)
+    }
+}
+
+impl<S, C> Incoming<C> for S
+where
+    S: Sized + Stream<Item = C>,
+    C: Channel,
+{
+}
--- a/tarpc/src/server/limits.rs
+++ b/tarpc/src/server/limits.rs
@@ -0,0 +1,5 @@
+/// Provides functionality to limit the number of active channels.
+pub mod channels_per_key;
+
+/// Provides a [channel](crate::server::Channel) that limits the number of in-flight requests.
+pub mod requests_per_channel;
--- a/tarpc/src/server/limits/channels_per_key.rs
+++ b/tarpc/src/server/limits/channels_per_key.rs
@@ -9,20 +9,23 @@ use crate::{
    util::Compact,
 };
 use fnv::FnvHashMap;
-use futures::{future::AbortRegistration, prelude::*, ready, stream::Fuse, task::*};
-use log::{debug, info, trace};
+use futures::{prelude::*, ready, stream::Fuse, task::*};
 use pin_project::pin_project;
 use std::sync::{Arc, Weak};
 use std::{
    collections::hash_map::Entry, convert::TryFrom, fmt, hash::Hash, marker::Unpin, pin::Pin,
-    time::SystemTime,
 };
 use tokio::sync::mpsc;
+use tracing::{debug, info, trace};

-/// A single-threaded filter that drops channels based on per-key limits.
+/// An [`Incoming`](crate::server::incoming::Incoming) stream that drops new channels based on
+/// per-key limits.
+///
+/// The decision to drop a Channel is made once at the time the Channel materializes. Once a
+/// Channel is yielded, it will not be prematurely dropped.
 #[pin_project]
 #[derive(Debug)]
-pub struct ChannelFilter<S, K, F>
+pub struct MaxChannelsPerKey<S, K, F>
 where
    K: Eq + Hash,
 {
@@ -35,7 +38,7 @@ where
    keymaker: F,
 }

-/// A channel that is tracked by a ChannelFilter.
+/// A channel that is tracked by [`MaxChannelsPerKey`].
 #[pin_project]
 #[derive(Debug)]
 pub struct TrackedChannel<C, K> {
@@ -103,6 +106,7 @@ where
 {
    type Req = C::Req;
    type Resp = C::Resp;
+    type Transport = C::Transport;

    fn config(&self) -> &server::Config {
        self.inner.config()
@@ -112,12 +116,8 @@ where
        self.inner.in_flight_requests()
    }

-    fn start_request(
-        mut self: Pin<&mut Self>,
-        id: u64,
-        deadline: SystemTime,
-    ) -> Result<AbortRegistration, super::in_flight_requests::AlreadyExistsError> {
-        self.inner_pin_mut().start_request(id, deadline)
+    fn transport(&self) -> &Self::Transport {
+        self.inner.transport()
    }
 }

@@ -133,7 +133,7 @@ impl<C, K> TrackedChannel<C, K> {
    }
 }

-impl<S, K, F> ChannelFilter<S, K, F>
+impl<S, K, F> MaxChannelsPerKey<S, K, F>
 where
    K: Eq + Hash,
    S: Stream,
@@ -142,7 +142,7 @@ where
    /// Sheds new channels to stay under configured limits.
    pub(crate) fn new(listener: S, channels_per_key: u32, keymaker: F) -> Self {
        let (dropped_keys_tx, dropped_keys) = mpsc::unbounded_channel();
-        ChannelFilter {
+        MaxChannelsPerKey {
            listener: listener.fuse(),
            channels_per_key,
            dropped_keys,
@@ -153,7 +153,7 @@ where
    }
 }

-impl<S, K, F> ChannelFilter<S, K, F>
+impl<S, K, F> MaxChannelsPerKey<S, K, F>
 where
    S: Stream,
    K: fmt::Display + Eq + Hash + Clone + Unpin,
@@ -171,11 +171,10 @@ where
        let tracker = self.as_mut().increment_channels_for_key(key.clone())?;

        trace!(
-            "[{}] Opening channel ({}/{}) channels for key.",
-            key,
-            Arc::strong_count(&tracker),
-            self.channels_per_key
-        );
+            channel_filter_key = %key,
+            open_channels = Arc::strong_count(&tracker),
+            max_open_channels = self.channels_per_key,
+            "Opening channel");

        Ok(TrackedChannel {
            tracker,
@@ -200,9 +199,10 @@ where
                let count = o.get().strong_count();
                if count >= TryFrom::try_from(*self_.channels_per_key).unwrap() {
                    info!(
-                        "[{}] Opened max channels from key ({}/{}).",
-                        key, count, self_.channels_per_key
-                    );
+                        channel_filter_key = %key,
+                        open_channels = count,
+                        max_open_channels = *self_.channels_per_key,
+                        "At open channel limit");
                    Err(key)
                } else {
                    Ok(o.get().upgrade().unwrap_or_else(|| {
@@ -233,7 +233,9 @@ where
        let self_ = self.project();
        match ready!(self_.dropped_keys.poll_recv(cx)) {
            Some(key) => {
-                debug!("All channels dropped for key [{}]", key);
+                debug!(
+                    channel_filter_key = %key,
+                    "All channels dropped");
                self_.key_counts.remove(&key);
                self_.key_counts.compact(0.1);
                Poll::Ready(())
@@ -243,7 +245,7 @@ where
    }
 }

-impl<S, K, F> Stream for ChannelFilter<S, K, F>
+impl<S, K, F> Stream for MaxChannelsPerKey<S, K, F>
 where
    S: Stream,
    K: fmt::Display + Eq + Hash + Clone + Unpin,
@@ -346,7 +348,7 @@ fn channel_filter_increment_channels_for_key() {
        key: &'static str,
    }
    let (_, listener) = futures::channel::mpsc::unbounded();
-    let filter = ChannelFilter::new(listener, 2, |chan: &TestChannel| chan.key);
+    let filter = MaxChannelsPerKey::new(listener, 2, |chan: &TestChannel| chan.key);
    pin_mut!(filter);
    let tracker1 = filter.as_mut().increment_channels_for_key("key").unwrap();
    assert_eq!(Arc::strong_count(&tracker1), 1);
@@ -367,7 +369,7 @@ fn channel_filter_handle_new_channel() {
        key: &'static str,
    }
    let (_, listener) = futures::channel::mpsc::unbounded();
-    let filter = ChannelFilter::new(listener, 2, |chan: &TestChannel| chan.key);
+    let filter = MaxChannelsPerKey::new(listener, 2, |chan: &TestChannel| chan.key);
    pin_mut!(filter);
    let channel1 = filter
        .as_mut()
@@ -399,7 +401,7 @@ fn channel_filter_poll_listener() {
        key: &'static str,
    }
    let (new_channels, listener) = futures::channel::mpsc::unbounded();
-    let filter = ChannelFilter::new(listener, 2, |chan: &TestChannel| chan.key);
+    let filter = MaxChannelsPerKey::new(listener, 2, |chan: &TestChannel| chan.key);
    pin_mut!(filter);

    new_channels
@@ -435,7 +437,7 @@ fn channel_filter_poll_closed_channels() {
        key: &'static str,
    }
    let (new_channels, listener) = futures::channel::mpsc::unbounded();
-    let filter = ChannelFilter::new(listener, 2, |chan: &TestChannel| chan.key);
+    let filter = MaxChannelsPerKey::new(listener, 2, |chan: &TestChannel| chan.key);
    pin_mut!(filter);

    new_channels
@@ -463,7 +465,7 @@ fn channel_filter_stream() {
        key: &'static str,
    }
    let (new_channels, listener) = futures::channel::mpsc::unbounded();
-    let filter = ChannelFilter::new(listener, 2, |chan: &TestChannel| chan.key);
+    let filter = MaxChannelsPerKey::new(listener, 2, |chan: &TestChannel| chan.key);
    pin_mut!(filter);

    new_channels
--- a/tarpc/src/server/limits/requests_per_channel.rs
+++ b/tarpc/src/server/limits/requests_per_channel.rs
@@ -0,0 +1,349 @@
+// Copyright 2020 Google LLC
+//
+// Use of this source code is governed by an MIT-style
+// license that can be found in the LICENSE file or at
+// https://opensource.org/licenses/MIT.
+
+use crate::{
+    server::{Channel, Config},
+    Response, ServerError,
+};
+use futures::{prelude::*, ready, task::*};
+use pin_project::pin_project;
+use std::{io, pin::Pin};
+
+/// A [`Channel`] that limits the number of concurrent requests by throttling.
+///
+/// Note that this is a very basic throttling heuristic. It is easy to set a number that is too low
+/// for the resources available to the server. For production use cases, a more advanced throttler
+/// is likely needed.
+#[pin_project]
+#[derive(Debug)]
+pub struct MaxRequests<C> {
+    max_in_flight_requests: usize,
+    #[pin]
+    inner: C,
+}
+
+impl<C> MaxRequests<C> {
+    /// Returns the inner channel.
+    pub fn get_ref(&self) -> &C {
+        &self.inner
+    }
+}
+
+impl<C> MaxRequests<C>
+where
+    C: Channel,
+{
+    /// Returns a new `MaxRequests` that wraps the given channel and limits concurrent requests to
+    /// `max_in_flight_requests`.
+    pub fn new(inner: C, max_in_flight_requests: usize) -> Self {
+        MaxRequests {
+            max_in_flight_requests,
+            inner,
+        }
+    }
+}
+
+impl<C> Stream for MaxRequests<C>
+where
+    C: Channel,
+{
+    type Item = <C as Stream>::Item;
+
+    fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context) -> Poll<Option<Self::Item>> {
+        while self.as_mut().in_flight_requests() >= *self.as_mut().project().max_in_flight_requests
+        {
+            ready!(self.as_mut().project().inner.poll_ready(cx)?);
+
+            match ready!(self.as_mut().project().inner.poll_next(cx)?) {
+                Some(r) => {
+                    let _entered = r.span.enter();
+                    tracing::info!(
+                        in_flight_requests = self.as_mut().in_flight_requests(),
+                        "ThrottleRequest",
+                    );
+
+                    self.as_mut().start_send(Response {
+                        request_id: r.request.id,
+                        message: Err(ServerError {
+                            kind: io::ErrorKind::WouldBlock,
+                            detail: "server throttled the request.".into(),
+                        }),
+                    })?;
+                }
+                None => return Poll::Ready(None),
+            }
+        }
+        self.project().inner.poll_next(cx)
+    }
+}
+
+impl<C> Sink<Response<<C as Channel>::Resp>> for MaxRequests<C>
+where
+    C: Channel,
+{
+    type Error = C::Error;
+
+    fn poll_ready(self: Pin<&mut Self>, cx: &mut Context) -> Poll<Result<(), Self::Error>> {
+        self.project().inner.poll_ready(cx)
+    }
+
+    fn start_send(
+        self: Pin<&mut Self>,
+        item: Response<<C as Channel>::Resp>,
+    ) -> Result<(), Self::Error> {
+        self.project().inner.start_send(item)
+    }
+
+    fn poll_flush(self: Pin<&mut Self>, cx: &mut Context) -> Poll<Result<(), Self::Error>> {
+        self.project().inner.poll_flush(cx)
+    }
+
+    fn poll_close(self: Pin<&mut Self>, cx: &mut Context) -> Poll<Result<(), Self::Error>> {
+        self.project().inner.poll_close(cx)
+    }
+}
+
+impl<C> AsRef<C> for MaxRequests<C> {
+    fn as_ref(&self) -> &C {
+        &self.inner
+    }
+}
+
+impl<C> Channel for MaxRequests<C>
+where
+    C: Channel,
+{
+    type Req = <C as Channel>::Req;
+    type Resp = <C as Channel>::Resp;
+    type Transport = <C as Channel>::Transport;
+
+    fn in_flight_requests(&self) -> usize {
+        self.inner.in_flight_requests()
+    }
+
+    fn config(&self) -> &Config {
+        self.inner.config()
+    }
+
+    fn transport(&self) -> &Self::Transport {
+        self.inner.transport()
+    }
+}
+
+/// An [`Incoming`](crate::server::incoming::Incoming) stream of channels that enforce limits on
+/// the number of in-flight requests.
+#[pin_project]
+#[derive(Debug)]
+pub struct MaxRequestsPerChannel<S> {
+    #[pin]
+    inner: S,
+    max_in_flight_requests: usize,
+}
+
+impl<S> MaxRequestsPerChannel<S>
+where
+    S: Stream,
+    <S as Stream>::Item: Channel,
+{
+    pub(crate) fn new(inner: S, max_in_flight_requests: usize) -> Self {
+        Self {
+            inner,
+            max_in_flight_requests,
+        }
+    }
+}
+
+impl<S> Stream for MaxRequestsPerChannel<S>
+where
+    S: Stream,
+    <S as Stream>::Item: Channel,
+{
+    type Item = MaxRequests<<S as Stream>::Item>;
+
+    fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context) -> Poll<Option<Self::Item>> {
+        match ready!(self.as_mut().project().inner.poll_next(cx)) {
+            Some(channel) => Poll::Ready(Some(MaxRequests::new(
+                channel,
+                *self.project().max_in_flight_requests,
+            ))),
+            None => Poll::Ready(None),
+        }
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    use crate::server::{
+        testing::{self, FakeChannel, PollExt},
+        TrackedRequest,
+    };
+    use pin_utils::pin_mut;
+    use std::{
+        marker::PhantomData,
+        time::{Duration, SystemTime},
+    };
+    use tracing::Span;
+
+    #[tokio::test]
+    async fn throttler_in_flight_requests() {
+        let throttler = MaxRequests {
+            max_in_flight_requests: 0,
+            inner: FakeChannel::default::<isize, isize>(),
+        };
+
+        pin_mut!(throttler);
+        for i in 0..5 {
+            throttler
+                .inner
+                .in_flight_requests
+                .start_request(
+                    i,
+                    SystemTime::now() + Duration::from_secs(1),
+                    Span::current(),
+                )
+                .unwrap();
+        }
+        assert_eq!(throttler.as_mut().in_flight_requests(), 5);
+    }
+
+    #[test]
+    fn throttler_poll_next_done() {
+        let throttler = MaxRequests {
+            max_in_flight_requests: 0,
+            inner: FakeChannel::default::<isize, isize>(),
+        };
+
+        pin_mut!(throttler);
+        assert!(throttler.as_mut().poll_next(&mut testing::cx()).is_done());
+    }
+
+    #[test]
+    fn throttler_poll_next_some() -> io::Result<()> {
+        let throttler = MaxRequests {
+            max_in_flight_requests: 1,
+            inner: FakeChannel::default::<isize, isize>(),
+        };
+
+        pin_mut!(throttler);
+        throttler.inner.push_req(0, 1);
+        assert!(throttler.as_mut().poll_ready(&mut testing::cx()).is_ready());
+        assert_eq!(
+            throttler
+                .as_mut()
+                .poll_next(&mut testing::cx())?
+                .map(|r| r.map(|r| (r.request.id, r.request.message))),
+            Poll::Ready(Some((0, 1)))
+        );
+        Ok(())
+    }
+
+    #[test]
+    fn throttler_poll_next_throttled() {
+        let throttler = MaxRequests {
+            max_in_flight_requests: 0,
+            inner: FakeChannel::default::<isize, isize>(),
+        };
+
+        pin_mut!(throttler);
+        throttler.inner.push_req(1, 1);
+        assert!(throttler.as_mut().poll_next(&mut testing::cx()).is_done());
+        assert_eq!(throttler.inner.sink.len(), 1);
+        let resp = throttler.inner.sink.get(0).unwrap();
+        assert_eq!(resp.request_id, 1);
+        assert!(resp.message.is_err());
+    }
+
+    #[test]
+    fn throttler_poll_next_throttled_sink_not_ready() {
+        let throttler = MaxRequests {
+            max_in_flight_requests: 0,
+            inner: PendingSink::default::<isize, isize>(),
+        };
+        pin_mut!(throttler);
+        assert!(throttler.poll_next(&mut testing::cx()).is_pending());
+
+        struct PendingSink<In, Out> {
+            ghost: PhantomData<fn(Out) -> In>,
+        }
+        impl PendingSink<(), ()> {
+            pub fn default<Req, Resp>(
+            ) -> PendingSink<io::Result<TrackedRequest<Req>>, Response<Resp>> {
+                PendingSink { ghost: PhantomData }
+            }
+        }
+        impl<In, Out> Stream for PendingSink<In, Out> {
+            type Item = In;
+            fn poll_next(self: Pin<&mut Self>, _: &mut Context) -> Poll<Option<Self::Item>> {
+                unimplemented!()
+            }
+        }
+        impl<In, Out> Sink<Out> for PendingSink<In, Out> {
+            type Error = io::Error;
+            fn poll_ready(self: Pin<&mut Self>, _: &mut Context) -> Poll<Result<(), Self::Error>> {
+                Poll::Pending
+            }
+            fn start_send(self: Pin<&mut Self>, _: Out) -> Result<(), Self::Error> {
+                Err(io::Error::from(io::ErrorKind::WouldBlock))
+            }
+            fn poll_flush(self: Pin<&mut Self>, _: &mut Context) -> Poll<Result<(), Self::Error>> {
+                Poll::Pending
+            }
+            fn poll_close(self: Pin<&mut Self>, _: &mut Context) -> Poll<Result<(), Self::Error>> {
+                Poll::Pending
+            }
+        }
+        impl<Req, Resp> Channel for PendingSink<io::Result<TrackedRequest<Req>>, Response<Resp>> {
+            type Req = Req;
+            type Resp = Resp;
+            type Transport = ();
+            fn config(&self) -> &Config {
+                unimplemented!()
+            }
+            fn in_flight_requests(&self) -> usize {
+                0
+            }
+            fn transport(&self) -> &() {
+                &()
+            }
+        }
+    }
+
+    #[tokio::test]
+    async fn throttler_start_send() {
+        let throttler = MaxRequests {
+            max_in_flight_requests: 0,
+            inner: FakeChannel::default::<isize, isize>(),
+        };
+
+        pin_mut!(throttler);
+        throttler
+            .inner
+            .in_flight_requests
+            .start_request(
+                0,
+                SystemTime::now() + Duration::from_secs(1),
+                Span::current(),
+            )
+            .unwrap();
+        throttler
+            .as_mut()
+            .start_send(Response {
+                request_id: 0,
+                message: Ok(1),
+            })
+            .unwrap();
+        assert_eq!(throttler.inner.in_flight_requests.len(), 0);
+        assert_eq!(
+            throttler.inner.sink.get(0),
+            Some(&Response {
+                request_id: 0,
+                message: Ok(1),
+            })
+        );
+    }
+}
--- a/tarpc/src/server/testing.rs
+++ b/tarpc/src/server/testing.rs
@@ -5,16 +5,15 @@
 // https://opensource.org/licenses/MIT.

 use crate::{
+    cancellations::{cancellations, CanceledRequests, RequestCancellation},
    context,
-    server::{Channel, Config},
+    server::{Channel, Config, ResponseGuard, TrackedRequest},
    Request, Response,
 };
-use futures::{future::AbortRegistration, task::*, Sink, Stream};
+use futures::{task::*, Sink, Stream};
 use pin_project::pin_project;
-use std::collections::VecDeque;
-use std::io;
-use std::pin::Pin;
-use std::time::SystemTime;
+use std::{collections::VecDeque, io, mem::ManuallyDrop, pin::Pin, time::SystemTime};
+use tracing::Span;

 #[pin_project]
 pub(crate) struct FakeChannel<In, Out> {
@@ -24,6 +23,8 @@ pub(crate) struct FakeChannel<In, Out> {
    pub sink: VecDeque<Out>,
    pub config: Config,
    pub in_flight_requests: super::in_flight_requests::InFlightRequests,
+    pub request_cancellation: RequestCancellation,
+    pub canceled_requests: CanceledRequests,
 }

 impl<In, Out> Stream for FakeChannel<In, Out>
@@ -64,12 +65,13 @@ impl<In, Resp> Sink<Response<Resp>> for FakeChannel<In, Response<Resp>> {
    }
 }

-impl<Req, Resp> Channel for FakeChannel<io::Result<Request<Req>>, Response<Resp>>
+impl<Req, Resp> Channel for FakeChannel<io::Result<TrackedRequest<Req>>, Response<Resp>>
 where
    Req: Unpin,
 {
    type Req = Req;
    type Resp = Resp;
+    type Transport = ();

    fn config(&self) -> &Config {
        &self.config
@@ -79,37 +81,44 @@ where
        self.in_flight_requests.len()
    }

-    fn start_request(
-        self: Pin<&mut Self>,
-        id: u64,
-        deadline: SystemTime,
-    ) -> Result<AbortRegistration, super::in_flight_requests::AlreadyExistsError> {
-        self.project()
-            .in_flight_requests
-            .start_request(id, deadline)
+    fn transport(&self) -> &() {
+        &()
    }
 }

-impl<Req, Resp> FakeChannel<io::Result<Request<Req>>, Response<Resp>> {
+impl<Req, Resp> FakeChannel<io::Result<TrackedRequest<Req>>, Response<Resp>> {
    pub fn push_req(&mut self, id: u64, message: Req) {
-        self.stream.push_back(Ok(Request {
-            context: context::Context {
-                deadline: SystemTime::UNIX_EPOCH,
-                trace_context: Default::default(),
+        let (_, abort_registration) = futures::future::AbortHandle::new_pair();
+        let (request_cancellation, _) = cancellations();
+        self.stream.push_back(Ok(TrackedRequest {
+            request: Request {
+                context: context::Context {
+                    deadline: SystemTime::UNIX_EPOCH,
+                    trace_context: Default::default(),
+                },
+                id,
+                message,
            },
-            id,
-            message,
+            abort_registration,
+            span: Span::none(),
+            response_guard: ManuallyDrop::new(ResponseGuard {
+                request_cancellation,
+                request_id: id,
+            }),
        }));
    }
 }

 impl FakeChannel<(), ()> {
-    pub fn default<Req, Resp>() -> FakeChannel<io::Result<Request<Req>>, Response<Resp>> {
+    pub fn default<Req, Resp>() -> FakeChannel<io::Result<TrackedRequest<Req>>, Response<Resp>> {
+        let (request_cancellation, canceled_requests) = cancellations();
        FakeChannel {
            stream: Default::default(),
            sink: Default::default(),
            config: Default::default(),
            in_flight_requests: Default::default(),
+            request_cancellation,
+            canceled_requests,
        }
    }
 }
--- a/tarpc/src/server/throttle.rs
+++ b/tarpc/src/server/throttle.rs
@@ -1,347 +0,0 @@
-// Copyright 2020 Google LLC
-//
-// Use of this source code is governed by an MIT-style
-// license that can be found in the LICENSE file or at
-// https://opensource.org/licenses/MIT.
-
-use super::{Channel, Config};
-use crate::{Response, ServerError};
-use futures::{future::AbortRegistration, prelude::*, ready, task::*};
-use log::debug;
-use pin_project::pin_project;
-use std::{io, pin::Pin, time::SystemTime};
-
-/// A [`Channel`] that limits the number of concurrent
-/// requests by throttling.
-#[pin_project]
-#[derive(Debug)]
-pub struct Throttler<C> {
-    max_in_flight_requests: usize,
-    #[pin]
-    inner: C,
-}
-
-impl<C> Throttler<C> {
-    /// Returns the inner channel.
-    pub fn get_ref(&self) -> &C {
-        &self.inner
-    }
-}
-
-impl<C> Throttler<C>
-where
-    C: Channel,
-{
-    /// Returns a new `Throttler` that wraps the given channel and limits concurrent requests to
-    /// `max_in_flight_requests`.
-    pub fn new(inner: C, max_in_flight_requests: usize) -> Self {
-        Throttler {
-            max_in_flight_requests,
-            inner,
-        }
-    }
-}
-
-impl<C> Stream for Throttler<C>
-where
-    C: Channel,
-{
-    type Item = <C as Stream>::Item;
-
-    fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context) -> Poll<Option<Self::Item>> {
-        while self.as_mut().in_flight_requests() >= *self.as_mut().project().max_in_flight_requests
-        {
-            ready!(self.as_mut().project().inner.poll_ready(cx)?);
-
-            match ready!(self.as_mut().project().inner.poll_next(cx)?) {
-                Some(request) => {
-                    debug!(
-                        "[{}] Client has reached in-flight request limit ({}/{}).",
-                        request.context.trace_id(),
-                        self.as_mut().in_flight_requests(),
-                        self.as_mut().project().max_in_flight_requests,
-                    );
-
-                    self.as_mut().start_send(Response {
-                        request_id: request.id,
-                        message: Err(ServerError {
-                            kind: io::ErrorKind::WouldBlock,
-                            detail: Some("Server throttled the request.".into()),
-                        }),
-                    })?;
-                }
-                None => return Poll::Ready(None),
-            }
-        }
-        self.project().inner.poll_next(cx)
-    }
-}
-
-impl<C> Sink<Response<<C as Channel>::Resp>> for Throttler<C>
-where
-    C: Channel,
-{
-    type Error = io::Error;
-
-    fn poll_ready(self: Pin<&mut Self>, cx: &mut Context) -> Poll<Result<(), Self::Error>> {
-        self.project().inner.poll_ready(cx)
-    }
-
-    fn start_send(self: Pin<&mut Self>, item: Response<<C as Channel>::Resp>) -> io::Result<()> {
-        self.project().inner.start_send(item)
-    }
-
-    fn poll_flush(self: Pin<&mut Self>, cx: &mut Context) -> Poll<io::Result<()>> {
-        self.project().inner.poll_flush(cx)
-    }
-
-    fn poll_close(self: Pin<&mut Self>, cx: &mut Context) -> Poll<io::Result<()>> {
-        self.project().inner.poll_close(cx)
-    }
-}
-
-impl<C> AsRef<C> for Throttler<C> {
-    fn as_ref(&self) -> &C {
-        &self.inner
-    }
-}
-
-impl<C> Channel for Throttler<C>
-where
-    C: Channel,
-{
-    type Req = <C as Channel>::Req;
-    type Resp = <C as Channel>::Resp;
-
-    fn in_flight_requests(&self) -> usize {
-        self.inner.in_flight_requests()
-    }
-
-    fn config(&self) -> &Config {
-        self.inner.config()
-    }
-
-    fn start_request(
-        self: Pin<&mut Self>,
-        id: u64,
-        deadline: SystemTime,
-    ) -> Result<AbortRegistration, super::in_flight_requests::AlreadyExistsError> {
-        self.project().inner.start_request(id, deadline)
-    }
-}
-
-/// A stream of throttling channels.
-#[pin_project]
-#[derive(Debug)]
-pub struct ThrottlerStream<S> {
-    #[pin]
-    inner: S,
-    max_in_flight_requests: usize,
-}
-
-impl<S> ThrottlerStream<S>
-where
-    S: Stream,
-    <S as Stream>::Item: Channel,
-{
-    pub(crate) fn new(inner: S, max_in_flight_requests: usize) -> Self {
-        Self {
-            inner,
-            max_in_flight_requests,
-        }
-    }
-}
-
-impl<S> Stream for ThrottlerStream<S>
-where
-    S: Stream,
-    <S as Stream>::Item: Channel,
-{
-    type Item = Throttler<<S as Stream>::Item>;
-
-    fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context) -> Poll<Option<Self::Item>> {
-        match ready!(self.as_mut().project().inner.poll_next(cx)) {
-            Some(channel) => Poll::Ready(Some(Throttler::new(
-                channel,
-                *self.project().max_in_flight_requests,
-            ))),
-            None => Poll::Ready(None),
-        }
-    }
-}
-
-#[cfg(test)]
-use super::testing::{self, FakeChannel, PollExt};
-#[cfg(test)]
-use crate::Request;
-#[cfg(test)]
-use pin_utils::pin_mut;
-#[cfg(test)]
-use std::{marker::PhantomData, time::Duration};
-
-#[tokio::test]
-async fn throttler_in_flight_requests() {
-    let throttler = Throttler {
-        max_in_flight_requests: 0,
-        inner: FakeChannel::default::<isize, isize>(),
-    };
-
-    pin_mut!(throttler);
-    for i in 0..5 {
-        throttler
-            .inner
-            .in_flight_requests
-            .start_request(i, SystemTime::now() + Duration::from_secs(1))
-            .unwrap();
-    }
-    assert_eq!(throttler.as_mut().in_flight_requests(), 5);
-}
-
-#[tokio::test]
-async fn throttler_start_request() {
-    let throttler = Throttler {
-        max_in_flight_requests: 0,
-        inner: FakeChannel::default::<isize, isize>(),
-    };
-
-    pin_mut!(throttler);
-    throttler
-        .as_mut()
-        .start_request(1, SystemTime::now() + Duration::from_secs(1))
-        .unwrap();
-    assert_eq!(throttler.inner.in_flight_requests.len(), 1);
-}
-
-#[test]
-fn throttler_poll_next_done() {
-    let throttler = Throttler {
-        max_in_flight_requests: 0,
-        inner: FakeChannel::default::<isize, isize>(),
-    };
-
-    pin_mut!(throttler);
-    assert!(throttler.as_mut().poll_next(&mut testing::cx()).is_done());
-}
-
-#[test]
-fn throttler_poll_next_some() -> io::Result<()> {
-    let throttler = Throttler {
-        max_in_flight_requests: 1,
-        inner: FakeChannel::default::<isize, isize>(),
-    };
-
-    pin_mut!(throttler);
-    throttler.inner.push_req(0, 1);
-    assert!(throttler.as_mut().poll_ready(&mut testing::cx()).is_ready());
-    assert_eq!(
-        throttler
-            .as_mut()
-            .poll_next(&mut testing::cx())?
-            .map(|r| r.map(|r| (r.id, r.message))),
-        Poll::Ready(Some((0, 1)))
-    );
-    Ok(())
-}
-
-#[test]
-fn throttler_poll_next_throttled() {
-    let throttler = Throttler {
-        max_in_flight_requests: 0,
-        inner: FakeChannel::default::<isize, isize>(),
-    };
-
-    pin_mut!(throttler);
-    throttler.inner.push_req(1, 1);
-    assert!(throttler.as_mut().poll_next(&mut testing::cx()).is_done());
-    assert_eq!(throttler.inner.sink.len(), 1);
-    let resp = throttler.inner.sink.get(0).unwrap();
-    assert_eq!(resp.request_id, 1);
-    assert!(resp.message.is_err());
-}
-
-#[test]
-fn throttler_poll_next_throttled_sink_not_ready() {
-    let throttler = Throttler {
-        max_in_flight_requests: 0,
-        inner: PendingSink::default::<isize, isize>(),
-    };
-    pin_mut!(throttler);
-    assert!(throttler.poll_next(&mut testing::cx()).is_pending());
-
-    struct PendingSink<In, Out> {
-        ghost: PhantomData<fn(Out) -> In>,
-    }
-    impl PendingSink<(), ()> {
-        pub fn default<Req, Resp>() -> PendingSink<io::Result<Request<Req>>, Response<Resp>> {
-            PendingSink { ghost: PhantomData }
-        }
-    }
-    impl<In, Out> Stream for PendingSink<In, Out> {
-        type Item = In;
-        fn poll_next(self: Pin<&mut Self>, _: &mut Context) -> Poll<Option<Self::Item>> {
-            unimplemented!()
-        }
-    }
-    impl<In, Out> Sink<Out> for PendingSink<In, Out> {
-        type Error = io::Error;
-        fn poll_ready(self: Pin<&mut Self>, _: &mut Context) -> Poll<Result<(), Self::Error>> {
-            Poll::Pending
-        }
-        fn start_send(self: Pin<&mut Self>, _: Out) -> Result<(), Self::Error> {
-            Err(io::Error::from(io::ErrorKind::WouldBlock))
-        }
-        fn poll_flush(self: Pin<&mut Self>, _: &mut Context) -> Poll<Result<(), Self::Error>> {
-            Poll::Pending
-        }
-        fn poll_close(self: Pin<&mut Self>, _: &mut Context) -> Poll<Result<(), Self::Error>> {
-            Poll::Pending
-        }
-    }
-    impl<Req, Resp> Channel for PendingSink<io::Result<Request<Req>>, Response<Resp>> {
-        type Req = Req;
-        type Resp = Resp;
-        fn config(&self) -> &Config {
-            unimplemented!()
-        }
-        fn in_flight_requests(&self) -> usize {
-            0
-        }
-        fn start_request(
-            self: Pin<&mut Self>,
-            _id: u64,
-            _deadline: SystemTime,
-        ) -> Result<AbortRegistration, super::in_flight_requests::AlreadyExistsError> {
-            unimplemented!()
-        }
-    }
-}
-
-#[tokio::test]
-async fn throttler_start_send() {
-    let throttler = Throttler {
-        max_in_flight_requests: 0,
-        inner: FakeChannel::default::<isize, isize>(),
-    };
-
-    pin_mut!(throttler);
-    throttler
-        .inner
-        .in_flight_requests
-        .start_request(0, SystemTime::now() + Duration::from_secs(1))
-        .unwrap();
-    throttler
-        .as_mut()
-        .start_send(Response {
-            request_id: 0,
-            message: Ok(1),
-        })
-        .unwrap();
-    assert_eq!(throttler.inner.in_flight_requests.len(), 0);
-    assert_eq!(
-        throttler.inner.sink.get(0),
-        Some(&Response {
-            request_id: 0,
-            message: Ok(1),
-        })
-    );
-}
--- a/tarpc/src/server/tokio.rs
+++ b/tarpc/src/server/tokio.rs
@@ -0,0 +1,113 @@
+use super::{Channel, Requests, Serve};
+use futures::{prelude::*, ready, task::*};
+use pin_project::pin_project;
+use std::pin::Pin;
+
+/// A future that drives the server by [spawning](tokio::spawn) a [`TokioChannelExecutor`](TokioChannelExecutor)
+/// for each new channel. Returned by
+/// [`Incoming::execute`](crate::server::incoming::Incoming::execute).
+#[must_use]
+#[pin_project]
+#[derive(Debug)]
+pub struct TokioServerExecutor<T, S> {
+    #[pin]
+    inner: T,
+    serve: S,
+}
+
+impl<T, S> TokioServerExecutor<T, S> {
+    pub(crate) fn new(inner: T, serve: S) -> Self {
+        Self { inner, serve }
+    }
+}
+
+/// A future that drives the server by [spawning](tokio::spawn) each [response
+/// handler](super::InFlightRequest::execute) on tokio's default executor. Returned by
+/// [`Channel::execute`](crate::server::Channel::execute).
+#[must_use]
+#[pin_project]
+#[derive(Debug)]
+pub struct TokioChannelExecutor<T, S> {
+    #[pin]
+    inner: T,
+    serve: S,
+}
+
+impl<T, S> TokioServerExecutor<T, S> {
+    fn inner_pin_mut<'a>(self: &'a mut Pin<&mut Self>) -> Pin<&'a mut T> {
+        self.as_mut().project().inner
+    }
+}
+
+impl<T, S> TokioChannelExecutor<T, S> {
+    fn inner_pin_mut<'a>(self: &'a mut Pin<&mut Self>) -> Pin<&'a mut T> {
+        self.as_mut().project().inner
+    }
+}
+
+// Send + 'static execution helper methods.
+
+impl<C> Requests<C>
+where
+    C: Channel,
+    C::Req: Send + 'static,
+    C::Resp: Send + 'static,
+{
+    /// Executes all requests using the given service function. Requests are handled concurrently
+    /// by [spawning](::tokio::spawn) each handler on tokio's default executor.
+    pub fn execute<S>(self, serve: S) -> TokioChannelExecutor<Self, S>
+    where
+        S: Serve<C::Req, Resp = C::Resp> + Send + 'static,
+    {
+        TokioChannelExecutor { inner: self, serve }
+    }
+}
+
+impl<St, C, Se> Future for TokioServerExecutor<St, Se>
+where
+    St: Sized + Stream<Item = C>,
+    C: Channel + Send + 'static,
+    C::Req: Send + 'static,
+    C::Resp: Send + 'static,
+    Se: Serve<C::Req, Resp = C::Resp> + Send + 'static + Clone,
+    Se::Fut: Send,
+{
+    type Output = ();
+
+    fn poll(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<()> {
+        while let Some(channel) = ready!(self.inner_pin_mut().poll_next(cx)) {
+            tokio::spawn(channel.execute(self.serve.clone()));
+        }
+        tracing::info!("Server shutting down.");
+        Poll::Ready(())
+    }
+}
+
+impl<C, S> Future for TokioChannelExecutor<Requests<C>, S>
+where
+    C: Channel + 'static,
+    C::Req: Send + 'static,
+    C::Resp: Send + 'static,
+    S: Serve<C::Req, Resp = C::Resp> + Send + 'static + Clone,
+    S::Fut: Send,
+{
+    type Output = ();
+
+    fn poll(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Self::Output> {
+        while let Some(response_handler) = ready!(self.inner_pin_mut().poll_next(cx)) {
+            match response_handler {
+                Ok(resp) => {
+                    let server = self.serve.clone();
+                    tokio::spawn(async move {
+                        resp.execute(server).await;
+                    });
+                }
+                Err(e) => {
+                    tracing::warn!("Requests stream errored out: {}", e);
+                    break;
+                }
+            }
+        }
+        Poll::Ready(())
+    }
+}
--- a/tarpc/src/trace.rs
+++ b/tarpc/src/trace.rs
@@ -16,11 +16,14 @@
 //! This crate's design is based on [opencensus
 //! tracing](https://opencensus.io/core-concepts/tracing/).

+use opentelemetry::trace::TraceContextExt;
 use rand::Rng;
 use std::{
+    convert::TryFrom,
    fmt::{self, Formatter},
-    mem,
+    num::{NonZeroU128, NonZeroU64},
 };
+use tracing_opentelemetry::OpenTelemetrySpanExt;

 /// A context for tracing the execution of processes, distributed or otherwise.
 ///
@@ -36,33 +39,50 @@ pub struct Context {
    /// before making an RPC, and the span ID is sent to the server. The server is free to create
    /// its own spans, for which it sets the client's span as the parent span.
    pub span_id: SpanId,
-    /// An identifier of the span that originated the current span. For example, if a server sends
-    /// an RPC in response to a client request that included a span, the server would create a span
-    /// for the RPC and set its parent to the span_id in the incoming request's context.
-    ///
-    /// If `parent_id` is `None`, then this is a root context.
-    pub parent_id: Option<SpanId>,
+    /// Indicates whether a sampler has already decided whether or not to sample the trace
+    /// associated with the Context. If `sampling_decision` is None, then a decision has not yet
+    /// been made. Downstream samplers do not need to abide by "no sample" decisions--for example,
+    /// an upstream client may choose to never sample, which may not make sense for the client's
+    /// dependencies. On the other hand, if an upstream process has chosen to sample this trace,
+    /// then the downstream samplers are expected to respect that decision and also sample the
+    /// trace. Otherwise, the full trace would not be able to be reconstructed.
+    pub sampling_decision: SamplingDecision,
 }

 /// A 128-bit UUID identifying a trace. All spans caused by the same originating span share the
 /// same trace ID.
-#[derive(Debug, Default, PartialEq, Eq, Hash, Clone, Copy)]
+#[derive(Default, PartialEq, Eq, Hash, Clone, Copy)]
 #[cfg_attr(feature = "serde1", derive(serde::Serialize, serde::Deserialize))]
-pub struct TraceId(u128);
+pub struct TraceId(#[cfg_attr(feature = "serde1", serde(with = "u128_serde"))] u128);

 /// A 64-bit identifier of a span within a trace. The identifier is unique within the span's trace.
-#[derive(Debug, Default, PartialEq, Eq, Hash, Clone, Copy)]
+#[derive(Default, PartialEq, Eq, Hash, Clone, Copy)]
 #[cfg_attr(feature = "serde1", derive(serde::Serialize, serde::Deserialize))]
 pub struct SpanId(u64);

+/// Indicates whether a sampler has decided whether or not to sample the trace associated with the
+/// Context. Downstream samplers do not need to abide by "no sample" decisions--for example, an
+/// upstream client may choose to never sample, which may not make sense for the client's
+/// dependencies. On the other hand, if an upstream process has chosen to sample this trace, then
+/// the downstream samplers are expected to respect that decision and also sample the trace.
+/// Otherwise, the full trace would not be able to be reconstructed reliably.
+#[derive(Debug, PartialEq, Eq, Hash, Clone, Copy)]
+#[cfg_attr(feature = "serde1", derive(serde::Serialize, serde::Deserialize))]
+#[repr(u8)]
+pub enum SamplingDecision {
+    /// The associated span was sampled by its creating process. Child spans must also be sampled.
+    Sampled,
+    /// The associated span was not sampled by its creating process.
+    Unsampled,
+}
+
 impl Context {
-    /// Constructs a new root context. A root context is one with no parent span.
-    pub fn new_root() -> Self {
-        let rng = &mut rand::thread_rng();
-        Context {
-            trace_id: TraceId::random(rng),
-            span_id: SpanId::random(rng),
-            parent_id: None,
+    /// Constructs a new context with the trace ID and sampling decision inherited from the parent.
+    pub(crate) fn new_child(&self) -> Self {
+        Self {
+            trace_id: self.trace_id,
+            span_id: SpanId::random(&mut rand::thread_rng()),
+            sampling_decision: self.sampling_decision,
        }
    }
 }
@@ -71,17 +91,128 @@ impl TraceId {
    /// Returns a random trace ID that can be assumed to be globally unique if `rng` generates
    /// actually-random numbers.
    pub fn random<R: Rng>(rng: &mut R) -> Self {
-        TraceId(u128::from(rng.next_u64()) << mem::size_of::<u64>() | u128::from(rng.next_u64()))
+        TraceId(rng.gen::<NonZeroU128>().get())
+    }
+
+    /// Returns true iff the trace ID is 0.
+    pub fn is_none(&self) -> bool {
+        self.0 == 0
    }
 }

 impl SpanId {
    /// Returns a random span ID that can be assumed to be unique within a single trace.
    pub fn random<R: Rng>(rng: &mut R) -> Self {
-        SpanId(rng.next_u64())
+        SpanId(rng.gen::<NonZeroU64>().get())
+    }
+
+    /// Returns true iff the span ID is 0.
+    pub fn is_none(&self) -> bool {
+        self.0 == 0
    }
 }

+impl From<TraceId> for u128 {
+    fn from(trace_id: TraceId) -> Self {
+        trace_id.0
+    }
+}
+
+impl From<u128> for TraceId {
+    fn from(trace_id: u128) -> Self {
+        Self(trace_id)
+    }
+}
+
+impl From<SpanId> for u64 {
+    fn from(span_id: SpanId) -> Self {
+        span_id.0
+    }
+}
+
+impl From<u64> for SpanId {
+    fn from(span_id: u64) -> Self {
+        Self(span_id)
+    }
+}
+
+impl From<opentelemetry::trace::TraceId> for TraceId {
+    fn from(trace_id: opentelemetry::trace::TraceId) -> Self {
+        Self::from(u128::from_be_bytes(trace_id.to_bytes()))
+    }
+}
+
+impl From<TraceId> for opentelemetry::trace::TraceId {
+    fn from(trace_id: TraceId) -> Self {
+        Self::from_bytes(u128::from(trace_id).to_be_bytes())
+    }
+}
+
+impl From<opentelemetry::trace::SpanId> for SpanId {
+    fn from(span_id: opentelemetry::trace::SpanId) -> Self {
+        Self::from(u64::from_be_bytes(span_id.to_bytes()))
+    }
+}
+
+impl From<SpanId> for opentelemetry::trace::SpanId {
+    fn from(span_id: SpanId) -> Self {
+        Self::from_bytes(u64::from(span_id).to_be_bytes())
+    }
+}
+
+impl TryFrom<&tracing::Span> for Context {
+    type Error = NoActiveSpan;
+
+    fn try_from(span: &tracing::Span) -> Result<Self, NoActiveSpan> {
+        let context = span.context();
+        if context.has_active_span() {
+            Ok(Self::from(context.span()))
+        } else {
+            Err(NoActiveSpan)
+        }
+    }
+}
+
+impl From<opentelemetry::trace::SpanRef<'_>> for Context {
+    fn from(span: opentelemetry::trace::SpanRef<'_>) -> Self {
+        let otel_ctx = span.span_context();
+        Self {
+            trace_id: TraceId::from(otel_ctx.trace_id()),
+            span_id: SpanId::from(otel_ctx.span_id()),
+            sampling_decision: SamplingDecision::from(otel_ctx),
+        }
+    }
+}
+
+impl From<SamplingDecision> for opentelemetry::trace::TraceFlags {
+    fn from(decision: SamplingDecision) -> Self {
+        match decision {
+            SamplingDecision::Sampled => opentelemetry::trace::TraceFlags::SAMPLED,
+            SamplingDecision::Unsampled => opentelemetry::trace::TraceFlags::default(),
+        }
+    }
+}
+
+impl From<&opentelemetry::trace::SpanContext> for SamplingDecision {
+    fn from(context: &opentelemetry::trace::SpanContext) -> Self {
+        if context.is_sampled() {
+            SamplingDecision::Sampled
+        } else {
+            SamplingDecision::Unsampled
+        }
+    }
+}
+
+impl Default for SamplingDecision {
+    fn default() -> Self {
+        Self::Unsampled
+    }
+}
+
+/// Returned when a [`Context`] cannot be constructed from a [`Span`](tracing::Span).
+#[derive(Debug)]
+pub struct NoActiveSpan;
+
 impl fmt::Display for TraceId {
    fn fmt(&self, f: &mut Formatter) -> Result<(), fmt::Error> {
        write!(f, "{:02x}", self.0)?;
@@ -89,9 +220,42 @@ impl fmt::Display for TraceId {
    }
 }

+impl fmt::Debug for TraceId {
+    fn fmt(&self, f: &mut Formatter) -> Result<(), fmt::Error> {
+        write!(f, "{:02x}", self.0)?;
+        Ok(())
+    }
+}
+
 impl fmt::Display for SpanId {
    fn fmt(&self, f: &mut Formatter) -> Result<(), fmt::Error> {
        write!(f, "{:02x}", self.0)?;
        Ok(())
    }
 }
+
+impl fmt::Debug for SpanId {
+    fn fmt(&self, f: &mut Formatter) -> Result<(), fmt::Error> {
+        write!(f, "{:02x}", self.0)?;
+        Ok(())
+    }
+}
+
+#[cfg(feature = "serde1")]
+mod u128_serde {
+    pub fn serialize<S>(u: &u128, serializer: S) -> Result<S::Ok, S::Error>
+    where
+        S: serde::Serializer,
+    {
+        serde::Serialize::serialize(&u.to_le_bytes(), serializer)
+    }
+
+    pub fn deserialize<'de, D>(deserializer: D) -> Result<u128, D::Error>
+    where
+        D: serde::Deserializer<'de>,
+    {
+        Ok(u128::from_le_bytes(serde::Deserialize::deserialize(
+            deserializer,
+        )?))
+    }
+}
--- a/tarpc/src/transport.rs
+++ b/tarpc/src/transport.rs
@@ -9,22 +9,32 @@
 //! The rpc crate is transport- and protocol-agnostic. Any transport that impls [`Transport`](sealed::Transport)
 //! can be plugged in, using whatever protocol it wants.

-use futures::prelude::*;
-use std::io;
-
 pub mod channel;

 pub(crate) mod sealed {
-    use super::*;
+    use futures::prelude::*;
+    use std::error::Error;

    /// A bidirectional stream ([`Sink`] + [`Stream`]) of messages.
-    pub trait Transport<SinkItem, Item>:
-        Stream<Item = io::Result<Item>> + Sink<SinkItem, Error = io::Error>
+    pub trait Transport<SinkItem, Item>
+    where
+        Self: Stream<Item = Result<Item, <Self as Sink<SinkItem>>::Error>>,
+        Self: Sink<SinkItem, Error = <Self as Transport<SinkItem, Item>>::TransportError>,
+        <Self as Sink<SinkItem>>::Error: Error,
    {
+        /// Associated type where clauses are not elaborated; this associated type allows users
+        /// bounding types by Transport to avoid having to explicitly add `T::Error: Error` to their
+        /// bounds.
+        type TransportError: Error + Send + Sync + 'static;
    }

-    impl<T, SinkItem, Item> Transport<SinkItem, Item> for T where
-        T: Stream<Item = io::Result<Item>> + Sink<SinkItem, Error = io::Error> + ?Sized
+    impl<T, SinkItem, Item, E> Transport<SinkItem, Item> for T
+    where
+        T: ?Sized,
+        T: Stream<Item = Result<Item, E>>,
+        T: Sink<SinkItem, Error = E>,
+        T::Error: Error + Send + Sync + 'static,
    {
+        type TransportError = E;
    }
 }
--- a/tarpc/src/transport/channel.rs
+++ b/tarpc/src/transport/channel.rs
@@ -6,13 +6,19 @@

 //! Transports backed by in-memory channels.

-use crate::PollIo;
 use futures::{task::*, Sink, Stream};
 use pin_project::pin_project;
-use std::io;
-use std::pin::Pin;
+use std::{error::Error, pin::Pin};
 use tokio::sync::mpsc;

+/// Errors that occur in the sending or receiving of messages over a channel.
+#[derive(thiserror::Error, Debug)]
+pub enum ChannelError {
+    /// An error occurred sending over the channel.
+    #[error("an error occurred sending over the channel")]
+    Send(#[source] Box<dyn Error + Send + Sync + 'static>),
+}
+
 /// Returns two unbounded channel peers. Each [`Stream`] yields items sent through the other's
 /// [`Sink`].
 pub fn unbounded<SinkItem, Item>() -> (
@@ -36,28 +42,33 @@ pub struct UnboundedChannel<Item, SinkItem> {
 }

 impl<Item, SinkItem> Stream for UnboundedChannel<Item, SinkItem> {
-    type Item = Result<Item, io::Error>;
+    type Item = Result<Item, ChannelError>;

-    fn poll_next(mut self: Pin<&mut Self>, cx: &mut Context<'_>) -> PollIo<Item> {
+    fn poll_next(
+        mut self: Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Option<Result<Item, ChannelError>>> {
        self.rx.poll_recv(cx).map(|option| option.map(Ok))
    }
 }

-impl<Item, SinkItem> Sink<SinkItem> for UnboundedChannel<Item, SinkItem> {
-    type Error = io::Error;
+const CLOSED_MESSAGE: &str = "the channel is closed and cannot accept new items for sending";

-    fn poll_ready(self: Pin<&mut Self>, _: &mut Context<'_>) -> Poll<io::Result<()>> {
+impl<Item, SinkItem> Sink<SinkItem> for UnboundedChannel<Item, SinkItem> {
+    type Error = ChannelError;
+
+    fn poll_ready(self: Pin<&mut Self>, _: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
        Poll::Ready(if self.tx.is_closed() {
-            Err(io::Error::from(io::ErrorKind::NotConnected))
+            Err(ChannelError::Send(CLOSED_MESSAGE.into()))
        } else {
            Ok(())
        })
    }

-    fn start_send(self: Pin<&mut Self>, item: SinkItem) -> io::Result<()> {
+    fn start_send(self: Pin<&mut Self>, item: SinkItem) -> Result<(), Self::Error> {
        self.tx
            .send(item)
-            .map_err(|_| io::Error::from(io::ErrorKind::NotConnected))
+            .map_err(|_| ChannelError::Send(CLOSED_MESSAGE.into()))
    }

    fn poll_flush(self: Pin<&mut Self>, _: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
@@ -65,7 +76,7 @@ impl<Item, SinkItem> Sink<SinkItem> for UnboundedChannel<Item, SinkItem> {
        Poll::Ready(Ok(()))
    }

-    fn poll_close(self: Pin<&mut Self>, _: &mut Context<'_>) -> Poll<io::Result<()>> {
+    fn poll_close(self: Pin<&mut Self>, _: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
        // UnboundedSender can't initiate closure.
        Poll::Ready(Ok(()))
    }
@@ -93,52 +104,45 @@ pub struct Channel<Item, SinkItem> {
 }

 impl<Item, SinkItem> Stream for Channel<Item, SinkItem> {
-    type Item = Result<Item, io::Error>;
+    type Item = Result<Item, ChannelError>;

-    fn poll_next(self: Pin<&mut Self>, cx: &mut Context<'_>) -> PollIo<Item> {
+    fn poll_next(
+        self: Pin<&mut Self>,
+        cx: &mut Context<'_>,
+    ) -> Poll<Option<Result<Item, ChannelError>>> {
        self.project().rx.poll_next(cx).map(|option| option.map(Ok))
    }
 }

 impl<Item, SinkItem> Sink<SinkItem> for Channel<Item, SinkItem> {
-    type Error = io::Error;
+    type Error = ChannelError;

-    fn poll_ready(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
+    fn poll_ready(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
        self.project()
            .tx
            .poll_ready(cx)
-            .map_err(convert_send_err_to_io)
+            .map_err(|e| ChannelError::Send(Box::new(e)))
    }

-    fn start_send(self: Pin<&mut Self>, item: SinkItem) -> io::Result<()> {
+    fn start_send(self: Pin<&mut Self>, item: SinkItem) -> Result<(), Self::Error> {
        self.project()
            .tx
            .start_send(item)
-            .map_err(convert_send_err_to_io)
+            .map_err(|e| ChannelError::Send(Box::new(e)))
    }

    fn poll_flush(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
        self.project()
            .tx
            .poll_flush(cx)
-            .map_err(convert_send_err_to_io)
+            .map_err(|e| ChannelError::Send(Box::new(e)))
    }

-    fn poll_close(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<io::Result<()>> {
+    fn poll_close(self: Pin<&mut Self>, cx: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
        self.project()
            .tx
            .poll_close(cx)
-            .map_err(convert_send_err_to_io)
-    }
-}
-
-fn convert_send_err_to_io(e: futures::channel::mpsc::SendError) -> io::Error {
-    if e.is_disconnected() {
-        io::Error::from(io::ErrorKind::NotConnected)
-    } else if e.is_full() {
-        io::Error::from(io::ErrorKind::WouldBlock)
-    } else {
-        io::Error::new(io::ErrorKind::Other, e)
+            .map_err(|e| ChannelError::Send(Box::new(e)))
    }
 }

@@ -147,17 +151,27 @@ fn convert_send_err_to_io(e: futures::channel::mpsc::SendError) -> io::Error {
 mod tests {
    use crate::{
        client, context,
-        server::{BaseChannel, Incoming},
-        transport,
+        server::{incoming::Incoming, BaseChannel},
+        transport::{
+            self,
+            channel::{Channel, UnboundedChannel},
+        },
    };
    use assert_matches::assert_matches;
    use futures::{prelude::*, stream};
-    use log::trace;
    use std::io;
+    use tracing::trace;
+
+    #[test]
+    fn ensure_is_transport() {
+        fn is_transport<SinkItem, Item, T: crate::Transport<SinkItem, Item>>() {}
+        is_transport::<(), (), UnboundedChannel<(), ()>>();
+        is_transport::<(), (), Channel<(), ()>>();
+    }

    #[tokio::test]
-    async fn integration() -> io::Result<()> {
-        let _ = env_logger::try_init();
+    async fn integration() -> anyhow::Result<()> {
+        let _ = tracing_subscriber::fmt::try_init();

        let (client_channel, server_channel) = transport::channel::unbounded();
        tokio::spawn(
@@ -167,16 +181,16 @@ mod tests {
                    future::ready(request.parse::<u64>().map_err(|_| {
                        io::Error::new(
                            io::ErrorKind::InvalidInput,
-                            format!("{:?} is not an int", request),
+                            format!("{request:?} is not an int"),
                        )
                    }))
                }),
        );

-        let client = client::new(client::Config::default(), client_channel).spawn()?;
+        let client = client::new(client::Config::default(), client_channel).spawn();

-        let response1 = client.call(context::current(), "123".into()).await?;
-        let response2 = client.call(context::current(), "abc".into()).await?;
+        let response1 = client.call(context::current(), "", "123".into()).await?;
+        let response2 = client.call(context::current(), "", "abc".into()).await?;

        trace!("response1: {:?}, response2: {:?}", response1, response2);

--- a/tarpc/src/util.rs
+++ b/tarpc/src/util.rs
@@ -38,11 +38,34 @@ where
    H: BuildHasher,
 {
    fn compact(&mut self, usage_ratio_threshold: f64) {
-        if self.capacity() > 1000 {
-            let usage_ratio = self.len() as f64 / self.capacity() as f64;
-            if usage_ratio < usage_ratio_threshold {
-                self.shrink_to_fit();
-            }
-        }
+        let usage_ratio_threshold = usage_ratio_threshold.clamp(f64::MIN_POSITIVE, 1.);
+        let cap = f64::max(1000., self.len() as f64 / usage_ratio_threshold);
+        self.shrink_to(cap as usize);
    }
 }
+
+#[test]
+fn test_compact() {
+    let mut map = HashMap::with_capacity(2048);
+    assert_eq!(map.capacity(), 3584);
+
+    // Make usage ratio 25%
+    for i in 0..896 {
+        map.insert(format!("k{i}"), "v");
+    }
+
+    map.compact(-1.0);
+    assert_eq!(map.capacity(), 3584);
+
+    map.compact(0.25);
+    assert_eq!(map.capacity(), 3584);
+
+    map.compact(0.50);
+    assert_eq!(map.capacity(), 1792);
+
+    map.compact(1.0);
+    assert_eq!(map.capacity(), 1792);
+
+    map.compact(2.0);
+    assert_eq!(map.capacity(), 1792);
+}
--- a/tarpc/src/util/serde.rs
+++ b/tarpc/src/util/serde.rs
@@ -5,31 +5,7 @@
 // https://opensource.org/licenses/MIT.

 use serde::{Deserialize, Deserializer, Serialize, Serializer};
-use std::{
-    io,
-    time::{Duration, SystemTime},
-};
-
-/// Serializes `system_time` as a `u64` equal to the number of seconds since the epoch.
-pub fn serialize_epoch_secs<S>(system_time: &SystemTime, serializer: S) -> Result<S::Ok, S::Error>
-where
-    S: Serializer,
-{
-    const ZERO_SECS: Duration = Duration::from_secs(0);
-    system_time
-        .duration_since(SystemTime::UNIX_EPOCH)
-        .unwrap_or(ZERO_SECS)
-        .as_secs() // Only care about second precision
-        .serialize(serializer)
-}
-
-/// Deserializes [`SystemTime`] from a `u64` equal to the number of seconds since the epoch.
-pub fn deserialize_epoch_secs<'de, D>(deserializer: D) -> Result<SystemTime, D::Error>
-where
-    D: Deserializer<'de>,
-{
-    Ok(SystemTime::UNIX_EPOCH + Duration::from_secs(u64::deserialize(deserializer)?))
-}
+use std::io;

 /// Serializes [`io::ErrorKind`] as a `u32`.
 #[allow(clippy::trivially_copy_pass_by_ref)] // Exact fn signature required by serde derive
--- a/tarpc/tests/compile_fail.rs
+++ b/tarpc/tests/compile_fail.rs
@@ -2,4 +2,8 @@
 fn ui() {
    let t = trybuild::TestCases::new();
    t.compile_fail("tests/compile_fail/*.rs");
+    #[cfg(feature = "tokio1")]
+    t.compile_fail("tests/compile_fail/tokio/*.rs");
+    #[cfg(all(feature = "serde-transport", feature = "tcp"))]
+    t.compile_fail("tests/compile_fail/serde_transport/*.rs");
 }
--- a/tarpc/tests/compile_fail/must_use_request_dispatch.rs
+++ b/tarpc/tests/compile_fail/must_use_request_dispatch.rs
@@ -0,0 +1,15 @@
+use tarpc::client;
+
+#[tarpc::service]
+trait World {
+    async fn hello(name: String) -> String;
+}
+
+fn main() {
+    let (client_transport, _) = tarpc::transport::channel::unbounded();
+
+    #[deny(unused_must_use)]
+    {
+        WorldClient::new(client::Config::default(), client_transport).dispatch;
+    }
+}
--- a/tarpc/tests/compile_fail/must_use_request_dispatch.stderr
+++ b/tarpc/tests/compile_fail/must_use_request_dispatch.stderr
@@ -0,0 +1,11 @@
+error: unused `RequestDispatch` that must be used
+  --> tests/compile_fail/must_use_request_dispatch.rs:13:9
+   |
+13 |         WorldClient::new(client::Config::default(), client_transport).dispatch;
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+   |
+note: the lint level is defined here
+  --> tests/compile_fail/must_use_request_dispatch.rs:11:12
+   |
+11 |     #[deny(unused_must_use)]
+   |            ^^^^^^^^^^^^^^^
--- a/tarpc/tests/compile_fail/serde_transport/must_use_tcp_connect.rs
+++ b/tarpc/tests/compile_fail/serde_transport/must_use_tcp_connect.rs
@@ -0,0 +1,9 @@
+use tarpc::serde_transport;
+use tokio_serde::formats::Json;
+
+fn main() {
+    #[deny(unused_must_use)]
+    {
+        serde_transport::tcp::connect::<_, (), (), _, _>("0.0.0.0:0", Json::default);
+    }
+}
--- a/tarpc/tests/compile_fail/serde_transport/must_use_tcp_connect.stderr
+++ b/tarpc/tests/compile_fail/serde_transport/must_use_tcp_connect.stderr
@@ -0,0 +1,11 @@
+error: unused `Connect` that must be used
+ --> tests/compile_fail/serde_transport/must_use_tcp_connect.rs:7:9
+  |
+7 |         serde_transport::tcp::connect::<_, (), (), _, _>("0.0.0.0:0", Json::default);
+  |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  |
+note: the lint level is defined here
+ --> tests/compile_fail/serde_transport/must_use_tcp_connect.rs:5:12
+  |
+5 |     #[deny(unused_must_use)]
+  |            ^^^^^^^^^^^^^^^
--- a/tarpc/tests/compile_fail/tarpc_server_missing_async.rs
+++ b/tarpc/tests/compile_fail/tarpc_server_missing_async.rs
@@ -7,8 +7,8 @@ struct HelloServer;

 #[tarpc::server]
 impl World for HelloServer {
-    fn hello(name: String) ->  String {
-        format!("Hello, {}!", name)
+    fn hello(name: String) -> String {
+        format!("Hello, {name}!", name)
    }
 }

--- a/tarpc/tests/compile_fail/tarpc_server_missing_async.stderr
+++ b/tarpc/tests/compile_fail/tarpc_server_missing_async.stderr
@@ -7,5 +7,5 @@ error: not all trait items implemented, missing: `HelloFut`
 error: hint: `#[tarpc::server]` only rewrites async fns, and `fn hello` is not async
  --> $DIR/tarpc_server_missing_async.rs:10:5
   |
-10 |     fn hello(name: String) ->  String {
+10 |     fn hello(name: String) -> String {
   |     ^^
--- a/tarpc/tests/compile_fail/tokio/must_use_channel_executor.rs
+++ b/tarpc/tests/compile_fail/tokio/must_use_channel_executor.rs
@@ -0,0 +1,29 @@
+use tarpc::{
+    context,
+    server::{self, Channel},
+};
+
+#[tarpc::service]
+trait World {
+    async fn hello(name: String) -> String;
+}
+
+#[derive(Clone)]
+struct HelloServer;
+
+#[tarpc::server]
+impl World for HelloServer {
+    async fn hello(self, _: context::Context, name: String) -> String {
+        format!("Hello, {name}!")
+    }
+}
+
+fn main() {
+    let (_, server_transport) = tarpc::transport::channel::unbounded();
+    let server = server::BaseChannel::with_defaults(server_transport);
+
+    #[deny(unused_must_use)]
+    {
+        server.execute(HelloServer.serve());
+    }
+}
--- a/tarpc/tests/compile_fail/tokio/must_use_channel_executor.stderr
+++ b/tarpc/tests/compile_fail/tokio/must_use_channel_executor.stderr
@@ -0,0 +1,11 @@
+error: unused `TokioChannelExecutor` that must be used
+  --> tests/compile_fail/tokio/must_use_channel_executor.rs:27:9
+   |
+27 |         server.execute(HelloServer.serve());
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+   |
+note: the lint level is defined here
+  --> tests/compile_fail/tokio/must_use_channel_executor.rs:25:12
+   |
+25 |     #[deny(unused_must_use)]
+   |            ^^^^^^^^^^^^^^^
--- a/tarpc/tests/compile_fail/tokio/must_use_server_executor.rs
+++ b/tarpc/tests/compile_fail/tokio/must_use_server_executor.rs
@@ -0,0 +1,30 @@
+use futures::stream::once;
+use tarpc::{
+    context,
+    server::{self, incoming::Incoming},
+};
+
+#[tarpc::service]
+trait World {
+    async fn hello(name: String) -> String;
+}
+
+#[derive(Clone)]
+struct HelloServer;
+
+#[tarpc::server]
+impl World for HelloServer {
+    async fn hello(self, _: context::Context, name: String) -> String {
+        format!("Hello, {name}!")
+    }
+}
+
+fn main() {
+    let (_, server_transport) = tarpc::transport::channel::unbounded();
+    let server = once(async move { server::BaseChannel::with_defaults(server_transport) });
+
+    #[deny(unused_must_use)]
+    {
+        server.execute(HelloServer.serve());
+    }
+}
--- a/tarpc/tests/compile_fail/tokio/must_use_server_executor.stderr
+++ b/tarpc/tests/compile_fail/tokio/must_use_server_executor.stderr
@@ -0,0 +1,11 @@
+error: unused `TokioServerExecutor` that must be used
+  --> tests/compile_fail/tokio/must_use_server_executor.rs:28:9
+   |
+28 |         server.execute(HelloServer.serve());
+   |         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+   |
+note: the lint level is defined here
+  --> tests/compile_fail/tokio/must_use_server_executor.rs:26:12
+   |
+26 |     #[deny(unused_must_use)]
+   |            ^^^^^^^^^^^^^^^
--- a/tarpc/tests/dataservice.rs
+++ b/tarpc/tests/dataservice.rs
@@ -1,9 +1,8 @@
 use futures::prelude::*;
-use std::io;
 use tarpc::serde_transport;
 use tarpc::{
    client, context,
-    server::{BaseChannel, Incoming},
+    server::{incoming::Incoming, BaseChannel},
 };
 use tokio_serde::formats::Json;

@@ -33,7 +32,7 @@ impl ColorProtocol for ColorServer {
 }

 #[tokio::test]
-async fn test_call() -> io::Result<()> {
+async fn test_call() -> anyhow::Result<()> {
    let transport = tarpc::serde_transport::tcp::listen("localhost:56797", Json::default).await?;
    let addr = transport.local_addr();
    tokio::spawn(
@@ -45,7 +44,7 @@ async fn test_call() -> io::Result<()> {
    );

    let transport = serde_transport::tcp::connect(addr, Json::default).await?;
-    let client = ColorProtocolClient::new(client::Config::default(), transport).spawn()?;
+    let client = ColorProtocolClient::new(client::Config::default(), transport).spawn();

    let color = client
        .get_opposite_color(context::current(), TestData::White)
--- a/tarpc/tests/service_functional.rs
+++ b/tarpc/tests/service_functional.rs
@@ -3,14 +3,11 @@ use futures::{
    future::{join_all, ready, Ready},
    prelude::*,
 };
-use std::{
-    io,
-    time::{Duration, SystemTime},
-};
+use std::time::{Duration, SystemTime};
 use tarpc::{
    client::{self},
    context,
-    server::{self, BaseChannel, Channel, Incoming},
+    server::{self, incoming::Incoming, BaseChannel, Channel},
    transport::channel,
 };
 use tokio::join;
@@ -34,13 +31,13 @@ impl Service for Server {
    type HeyFut = Ready<String>;

    fn hey(self, _: context::Context, name: String) -> Self::HeyFut {
-        ready(format!("Hey, {}.", name))
+        ready(format!("Hey, {name}."))
    }
 }

 #[tokio::test]
-async fn sequential() -> io::Result<()> {
-    let _ = env_logger::try_init();
+async fn sequential() -> anyhow::Result<()> {
+    let _ = tracing_subscriber::fmt::try_init();

    let (tx, rx) = channel::unbounded();

@@ -50,7 +47,7 @@ async fn sequential() -> io::Result<()> {
            .execute(Server.serve()),
    );

-    let client = ServiceClient::new(client::Config::default(), tx).spawn()?;
+    let client = ServiceClient::new(client::Config::default(), tx).spawn();

    assert_matches!(client.add(context::current(), 1, 2).await, Ok(3));
    assert_matches!(
@@ -61,7 +58,7 @@ async fn sequential() -> io::Result<()> {
 }

 #[tokio::test]
-async fn dropped_channel_aborts_in_flight_requests() -> io::Result<()> {
+async fn dropped_channel_aborts_in_flight_requests() -> anyhow::Result<()> {
    #[tarpc_plugins::service]
    trait Loop {
        async fn r#loop();
@@ -82,16 +79,14 @@ async fn dropped_channel_aborts_in_flight_requests() -> io::Result<()> {
        }
    }

-    let _ = env_logger::try_init();
+    let _ = tracing_subscriber::fmt::try_init();

    let (tx, rx) = channel::unbounded();

    // Set up a client that initiates a long-lived request.
    // The request will complete in error when the server drops the connection.
    tokio::spawn(async move {
-        let client = LoopClient::new(client::Config::default(), tx)
-            .spawn()
-            .unwrap();
+        let client = LoopClient::new(client::Config::default(), tx).spawn();

        let mut ctx = context::current();
        ctx.deadline = SystemTime::now() + Duration::from_secs(60 * 60);
@@ -113,11 +108,11 @@ async fn dropped_channel_aborts_in_flight_requests() -> io::Result<()> {

 #[cfg(all(feature = "serde-transport", feature = "tcp"))]
 #[tokio::test]
-async fn serde() -> io::Result<()> {
+async fn serde() -> anyhow::Result<()> {
    use tarpc::serde_transport;
    use tokio_serde::formats::Json;

-    let _ = env_logger::try_init();
+    let _ = tracing_subscriber::fmt::try_init();

    let transport = tarpc::serde_transport::tcp::listen("localhost:56789", Json::default).await?;
    let addr = transport.local_addr();
@@ -130,7 +125,7 @@ async fn serde() -> io::Result<()> {
    );

    let transport = serde_transport::tcp::connect(addr, Json::default).await?;
-    let client = ServiceClient::new(client::Config::default(), transport).spawn()?;
+    let client = ServiceClient::new(client::Config::default(), transport).spawn();

    assert_matches!(client.add(context::current(), 1, 2).await, Ok(3));
    assert_matches!(
@@ -142,8 +137,8 @@ async fn serde() -> io::Result<()> {
 }

 #[tokio::test]
-async fn concurrent() -> io::Result<()> {
-    let _ = env_logger::try_init();
+async fn concurrent() -> anyhow::Result<()> {
+    let _ = tracing_subscriber::fmt::try_init();

    let (tx, rx) = channel::unbounded();
    tokio::spawn(
@@ -152,7 +147,7 @@ async fn concurrent() -> io::Result<()> {
            .execute(Server.serve()),
    );

-    let client = ServiceClient::new(client::Config::default(), tx).spawn()?;
+    let client = ServiceClient::new(client::Config::default(), tx).spawn();

    let req1 = client.add(context::current(), 1, 2);
    let req2 = client.add(context::current(), 3, 4);
@@ -166,8 +161,8 @@ async fn concurrent() -> io::Result<()> {
 }

 #[tokio::test]
-async fn concurrent_join() -> io::Result<()> {
-    let _ = env_logger::try_init();
+async fn concurrent_join() -> anyhow::Result<()> {
+    let _ = tracing_subscriber::fmt::try_init();

    let (tx, rx) = channel::unbounded();
    tokio::spawn(
@@ -176,7 +171,7 @@ async fn concurrent_join() -> io::Result<()> {
            .execute(Server.serve()),
    );

-    let client = ServiceClient::new(client::Config::default(), tx).spawn()?;
+    let client = ServiceClient::new(client::Config::default(), tx).spawn();

    let req1 = client.add(context::current(), 1, 2);
    let req2 = client.add(context::current(), 3, 4);
@@ -191,8 +186,8 @@ async fn concurrent_join() -> io::Result<()> {
 }

 #[tokio::test]
-async fn concurrent_join_all() -> io::Result<()> {
-    let _ = env_logger::try_init();
+async fn concurrent_join_all() -> anyhow::Result<()> {
+    let _ = tracing_subscriber::fmt::try_init();

    let (tx, rx) = channel::unbounded();
    tokio::spawn(
@@ -201,7 +196,7 @@ async fn concurrent_join_all() -> io::Result<()> {
            .execute(Server.serve()),
    );

-    let client = ServiceClient::new(client::Config::default(), tx).spawn()?;
+    let client = ServiceClient::new(client::Config::default(), tx).spawn();

    let req1 = client.add(context::current(), 1, 2);
    let req2 = client.add(context::current(), 3, 4);
@@ -214,7 +209,7 @@ async fn concurrent_join_all() -> io::Result<()> {
 }

 #[tokio::test]
-async fn counter() -> io::Result<()> {
+async fn counter() -> anyhow::Result<()> {
    #[tarpc::service]
    trait Counter {
        async fn count() -> u32;
@@ -241,7 +236,7 @@ async fn counter() -> io::Result<()> {
        }
    });

-    let client = CounterClient::new(client::Config::default(), tx).spawn()?;
+    let client = CounterClient::new(client::Config::default(), tx).spawn();
    assert_matches!(client.count(context::current()).await, Ok(1));
    assert_matches!(client.count(context::current()).await, Ok(2));
Author	SHA1	Message	Date
Tim Kuehn	99bf3e62a3	Prepare release of 0.30.0	2022-08-12 16:08:33 -07:00
Tim Kuehn	68863e3db0	Remove Channel::request_cancellation. This trait fn returns a private type, which means it's useless for anyone using the Channel. Instead, add an inert (now-public) ResponseGuard to TrackedRequest that, when taken out of the ManuallyDrop, ensures a Channel's request state is cleaned up. It's preferable to make ResponseGuard public instead of RequestCancellations because it's a smaller API surface (no public methods, just a Drop fn) and harder to misuse, because it is already associated with the correct request ID to cancel.	2022-08-12 16:08:33 -07:00
Tim Kuehn	453ba1c074	Lower log level of log in the RPC callpath	2022-08-12 09:04:47 -07:00
Makro	e3eac1b4f5	Add LICENSE files to crates (#372 )	2022-08-10 17:11:50 -07:00
kkharji	0e102288a5	feat: re-export used packages (#371 ) ## Problem Library users might get stuck with or ran into issues while using tarpc because of incompatible third party libraries. in particular, tokio_serde and tokio_util. ## Solution This PR does the following: 1. re-export tokio_serde as part of feature serde-transport, because the end user imports it to use some serde-transport APIs. 2. Update third library packages to latest release and fix resulting issues from that. ## Important Notes tokio_util 7.3 DelayQueue::poll_expired API changed [0] therefore, InFlightRequests::poll_expired now returns Poll<Option<u64>> [0] https://docs.rs/tokio-util/latest/tokio_util/time/delay_queue/struct.DelayQueue.html#method.poll_expired	2022-07-15 10:14:49 -07:00
Tim Kuehn	4c8ba41b2f	#[allow(unstable_name_collisions)] for .ready() .ready() is being added to std, but in the meantime, I don't want to stop using PollTest::ready.	2022-06-07 01:29:14 -07:00
Tim Kuehn	946c627579	Remove unused field	2022-06-07 01:29:14 -07:00
Tim Kuehn	104dd71bba	Clean up Channel request data more reliably. When an InFlightRequest is dropped before response completion, request data in the Channel persists until either the request expires or the client cancels the request. In rare cases, requests with very large deadlines could clog up the Channel long after request processing ceases. This commit adds a drop hook to InFlightRequest so that if it is dropped before execution completes, a cancellation message is sent to the Channel so that it can clean up the associated request data. This only works for when using `InFlightRequest::execute` or `Channel::execute`. However, users of raw `Channel` have access to the `RequestCancellation` handle via `Channel::request_cancellation`, so they can implement a similar method if they wish to manually clean up request data. Note that once a Channel's request data is cleaned up, that request can never be responded to, even if a response is produced afterward. Fixes https://github.com/google/tarpc/issues/314	2022-06-07 01:29:04 -07:00
Tim Kuehn	012c481861	Move cancellation types into a dedicated module. Cancellation utilities could be useful for both client and server code.	2022-06-05 18:54:52 -07:00
Tim Kuehn	dc12bd09aa	Annotate types that impl Future with #[must_use]. These types do nothing unless polled / .awaited. Annotating them with #[must_use] helps prevent a common class of coding errors. Fixes https://github.com/google/tarpc/issues/368.	2022-06-05 18:54:52 -07:00
Tim Kuehn	2594ea8ce9	Prepare release of 0.29.0	2022-06-05 15:26:33 -07:00
Tim Kuehn	839b87e394	Serialize RPC deadline as a Duration. Duration was previously serialized as SystemTime. However, absolute times run into problems with clock skew: if the remote machine's clock is too far in the future, the RPC deadline will be exceeded before request processing can begin. Conversely, if the remote machine's clock is too far in the past, the RPC deadline will not be enforced. By converting the absolute deadline to a relative duration, clock skew is no longer relevant, as the remote machine will convert the deadline into a time relative to its own clock. This mirrors how the gRPC HTTP2 protocol includes a Timeout in the request headers [0] but the SDK uses timestamps [1]. Keeping the absolute time in the core APIs maintains all the benefits of today, namely, natural deadline propagation between RPC hops when using the current context. This serialization strategy means that, generally, the remote machine's deadline will be slightly in the future compared to the local machine. Depending on network transfer latencies, this could be microseconds to milliseconds, or worse in the worst case. Because the deadline is not intended for high-precision scenarios, I don't view this is as problematic. Because this change only affects the serialization layer, local transports that bypass serialization are not affected. [0] https://github.com/grpc/grpc/blob/master/doc/PROTOCOL-HTTP2.md [1] https://grpc.io/blog/deadlines/#setting-a-deadline	2022-05-26 15:18:49 -07:00
Tim Kuehn	57d0638a99	Add rpc.deadline tag to Opentelemetry traces.	2022-05-26 15:18:49 -07:00
Tim Kuehn	a3a6404a30	Prepare release of 0.28.0	2022-04-06 22:07:07 -07:00
Tim Kuehn	b36eac80b1	Bump minimum rust version to 1.58.0	2022-04-06 21:53:56 -07:00
Bruno	d7070e4bc3	Update opentelemetry and related dependencies (#362 )	2022-04-03 14:09:14 -07:00
Tim Kuehn	b5d1828308	Use captured identifiers in format strings. This was stabilized in Rust 1.58.0: https://blog.rust-lang.org/2022/01/13/Rust-1.58.0.html	2022-01-13 15:00:44 -08:00
Zak Cutner	92cfe63c4f	Use single-threaded Tokio runtime (#360 )	2022-01-06 20:59:51 -08:00
Tim Kuehn	839a2f067c	Update example to latest version of Clap	2021-12-27 22:56:17 -08:00
David Kleingeld	b5d593488c	Derive more traits for RpcError (#359 ) Makes RpcError derive Clone, PartialEq, Eq, Hash, Serialize and Deserialize.	2021-12-27 22:00:58 -08:00
Tim Kuehn	eea38b8bf4	Simplify code with const assert!. The code that prevents compilation on systems where usize is larger than u64 previously used a const index-out-of-bounds trick. That code can now be replaced with assert!, as const panic! has landed in 1.57.0 stable.	2021-12-03 15:20:33 -08:00
Shi Yan	70493c15f4	Fix a compiling issue of the official example (#358 ) Fix a compiling issue of the official example because of the following error : ``` error[E0599]: the method `execute` exists for struct `BaseChannel<_, _, UnboundedChannel<ClientMessage<_>, Response<_>>>`, but its trait bounds were not satisfied --> src/main.rs:39:25 \| 39 \| tokio::spawn(server.execute(HelloServer.serve())); \| ^^^^^^^ method cannot be called on `BaseChannel<_, _, UnboundedChannel<ClientMessage<_>, Response<_>>>` due to unsatisfied trait bounds \| = note: the following trait bounds were not satisfied: `<&BaseChannel<_, _, UnboundedChannel<ClientMessage<_>, Response<_>>> as futures::Stream>::Item = _` which is required by `&BaseChannel<_, _, UnboundedChannel<ClientMessage<_>, Response<_>>>: tarpc::server::incoming::Incoming<_>` `&BaseChannel<_, _, UnboundedChannel<ClientMessage<_>, Response<_>>>: futures::Stream` which is required by `&BaseChannel<_, _, UnboundedChannel<ClientMessage<_>, Response<_>>>: tarpc::server::incoming::Incoming<_>` = help: items from traits can only be used if the trait is in scope help: the following trait is implemented but not in scope; perhaps add a `use` for it: \| 1 \| use tarpc::server::Channel; \| ``` See https://github.com/google/tarpc/pull/358#issuecomment-981953193 for the root cause.	2021-11-29 17:01:16 -08:00
baptiste0928	f7c5d6a7c3	Fix example-service (#355 ) Fixes the compilation of the example-service crate (the Clap trait has been renamed Parser in clap-rs/clap@d840d56).	2021-11-15 08:39:04 -08:00
Scott Kirkpatrick	98c5d2a18b	Re-add typo fixes (#353 ) The typo fixes that were added by commit b5d9aaa were accidentally reverted by commit `1e680e3`, this will add them back	2021-11-08 10:07:21 -08:00
Tim Kuehn	46b534f7c6	Use HashMap::shrink_to in impl of Comapct::compact.	2021-10-21 17:03:57 -07:00
Tim Kuehn	42b4fc52b1	Set rust-version to 1.56	2021-10-21 16:08:15 -07:00
Tim Kuehn	350dbcdad0	Upgrade to Rust 2021!	2021-10-21 14:10:21 -07:00
Tim Kuehn	b1b4461d89	Prepare release of 0.27.2	2021-10-08 22:31:56 -07:00
Tim Kuehn	f694b7573a	Close TcpStream when client disconnects. An attempt at a clean shutdown helps the server to drop its connections more quickly. Testing this uncovered a latent bug in DelayQueue wherein `poll_expired` yields `Pending` when empty. A workaround was added to `InFlightRequests::poll_expired`: check if there are actually any outstanding requests before calling `DelayQueue::poll_expired`.	2021-10-08 22:13:24 -07:00
Tim Kuehn	1e680e3a5a	Fix typos in docs. Fixes https://github.com/google/tarpc/issues/352.	2021-10-08 19:19:50 -07:00
Tim Kuehn	2591d21e94	Update release notes to mention io::Error =	2021-09-23 13:57:43 -07:00
Tim Kuehn	6632f68d95	Prepare for 0.27 release	2021-09-22 15:41:34 -07:00
Dmitry Kakurin	25985ad56a	Update README.md (#350 ) Fixed 2 typos	2021-09-01 17:58:49 -07:00
Tim Kuehn	d6a24e9420	Address Clippy lint	2021-08-24 12:40:18 -07:00
Tim Kuehn	281a78f3c7	Add tokio-serde-bincode feature	2021-08-24 12:37:57 -07:00
Julian Tescher	a0787d0091	Update to opentelemetry 0.16.x (#349 )	2021-08-17 00:00:07 -04:00
Frederik-Baetens	d2acba0e8a	add serde-transport-json feature flag (#346 ) In general, it should be possible to use, or at least import all functionality of a library, when having only that library in your cargo.toml.	2021-05-06 08:41:57 -07:00
Tim Kuehn	ea7b6763c4	Refactor server module. In the interest of the user's attention, some ancillary APIs have been moved to new submodules: - server::limits contains what was previously called Throttler and ChannelFilter. Both of those names were very generic, when the methods applied by these types were very specific (and also simplistic). Renames have occurred: - ThrottlerStream => MaxRequestsPerChannel - Throttler => MaxRequests - ChannelFilter => MaxChannelsPerKey - server::incoming contains the Incoming trait. - server::tokio contains the tokio-specific helper types. The 5 structs and 1 enum remaining in the base server module are all core to the functioning of the server.	2021-04-21 17:05:49 -07:00
Tim Kuehn	eb67c540b9	Use more structured errors in client.	2021-04-21 14:54:45 -07:00
Tim Kuehn	4151d0abd3	Move Span creation into BaseChannel. It's important for Channel decorators, like Throttler, to have access to the Span. This means that the BaseChannel becomes responsible for starting its own requests. Actually, this simplifies the integration for the Channel users, as they can assume any yielded requests are already tracked. This entails the following breaking changes: - removed trait method Channel::start_request as it is now done internally.	2021-04-21 14:54:45 -07:00
Tim Kuehn	d0c11a6efa	Change RPC error type from io::Error => RpcError. Becaue tarpc is a library, not an application, it should strive to use structured errors in its API so that users have maximal flexibility in how they handle errors. io::Error makes that hard, because it is a kitchen-sink error type. RPCs in particular only have 3 classes of errors: - The connection breaks. - The request expires. - The server decides not to process the request. (Of course, RPCs themselves can have application-specific errors, but from the perspective of the RPC library, those can be classified as successful responsees).	2021-04-20 18:29:55 -07:00
Tim Kuehn	82c4da1743	Prepare release of v0.26.2	2021-04-20 11:28:15 -07:00
Tim Kuehn	0a15e0b75c	Rustdoc: link RPC futures to their methods.	2021-04-20 11:25:26 -07:00
Tim Kuehn	0b315c29bf	It's not currently possible to document the enum variants, which means projects that #[deny(missing_docs)] wouldn't compile if using tarpc services.	2021-04-20 09:01:39 -07:00
Tim Kuehn	56f09bf61f	Fix log that's split across lines.	2021-04-17 17:15:16 -07:00
Tim Kuehn	6d82e82419	Fix formatting	2021-04-16 16:51:21 -07:00
Tim Kuehn	9bebaf814a	Address clippy lint	2021-04-14 17:49:27 -07:00
Tim Kuehn	5f4d6e6008	Prepare release of v0.26.0	2021-04-14 17:08:44 -07:00
Tim Kuehn	07d07d7ba3	Remove tracing_appender as it does not support build target mipsel-unknown-linux-gnu	2021-04-01 19:37:02 -07:00
Tim Kuehn	a41bbf65b2	Use rustfmt instead of cargo fmt so that diff is only printed once	2021-04-01 17:24:34 -07:00
Tim Kuehn	21e2f7ca62	Tear out requirement that Transport's error type is io::Error.	2021-04-01 17:24:34 -07:00
Tim Kuehn	7b7c182411	Instrument tarpc with tracing. tarpc is now instrumented with tracing primitives extended with OpenTelemetry traces. Using a compatible tracing-opentelemetry subscriber like Jaeger, each RPC can be traced through the client, server, amd other dependencies downstream of the server. Even for applications not connected to a distributed tracing collector, the instrumentation can also be ingested by regular loggers like env_logger. # Breaking Changes ## Logging Logged events are now structured using tracing. For applications using a logger and not a tracing subscriber, these logs may look different or contain information in a less consumable manner. The easiest solution is to add a tracing subscriber that logs to stdout, such as tracing_subscriber::fmt. ## Context - Context no longer has parent_span, which was actually never needed, because the context sent in an RPC is inherently the parent context. For purposes of distributed tracing, the client side of the RPC has all necessary information to link the span to its parent; the server side need do nothing more than export the (trace ID, span ID) tuple. - Context has a new field, SamplingDecision, which has two variants, Sampled and Unsampled. This field can be used by downstream systems to determine whether a trace needs to be exported. If the parent span is sampled, the expectation is that all child spans be exported, as well; to do otherwise could result in lossy traces being exported. Note that if an Openetelemetry tracing subscriber is not installed, the fallback context will still be used, but the Context's sampling decision will always be inherited by the parent Context's sampling decision. - Context::scope has been removed. Context propagation is now done via tracing's task-local spans. Spans can be propagated across tasks via Span::in_scope. When a service receives a request, it attaches an Opentelemetry context to the local Span created before request handling, and this context contains the request deadline. This span-local deadline is retrieved by Context::current, but it cannot be modified so that future Context::current calls contain a different deadline. However, the deadline in the context passed into an RPC call will override it, so users can retrieve the current context and then modify the deadline field, as has been historically possible. - Context propgation precedence changes: when an RPC is initiated, the current Span's Opentelemetry context takes precedence over the trace context passed into the RPC method. If there is no current Span, then the trace context argument is used as it has been historically. Note that Opentelemetry context propagation requires an Opentelemetry tracing subscriber to be installed. ## Server - The server::Channel trait now has an additional required associated type and method which returns the underlying transport. This makes it more ergonomic for users to retrieve transport-specific information, like IP Address. BaseChannel implements Channel::transport by returning the underlying transport, and channel decorators like Throttler just delegate to the Channel::transport method of the wrapped channel. # References [1] https://github.com/tokio-rs/tracing [2] https://opentelemetry.io [3] https://github.com/open-telemetry/opentelemetry-rust/tree/main/opentelemetry-jaeger [4] https://github.com/env-logger-rs/env_logger	2021-04-01 17:24:34 -07:00
Ben Ludewig	db0c778ead	Serialize u128 TraceId as LE bytes (#344 )	2021-03-30 08:41:19 -07:00
Tim Kuehn	c3efb83ac1	Add more context to errors returned by serde transport	2021-03-28 20:03:03 -07:00
Tim Kuehn	3d7b0171fe	Fix cargo fmt portion of pre-commit	2021-03-26 19:39:56 -07:00
oblique	c191ff5b2e	Do not enable tokio-serde/json by default (#345 )	2021-03-26 18:22:44 -07:00
Tim Kuehn	90bc7f741d	Fix up imports	2021-03-17 12:44:39 -07:00
Kitsu	d3f6c01df2	Reduce required tokio features (#343 ) * Move async tests behind cfg-ed mod * Use explicit tokio features for the example * Use only relative crate path for example dependency	2021-03-17 12:30:18 -07:00
Tim Kuehn	c6450521e6	Add method to run a future in the current context. Previously, `Context::current` would always return a new context. Now, it uses tokio task-local data to look for the current context. Tokio task locals are not actually tied to a tokio executor; instead, they provide data scoped to a future. The basic pattern is: ```rust let ctx = Context::new_root(); ctx.scope(async { let ctx2 = context::current(); assert_eq!(ctx2.trace_context.span_id, ctx.trace_context.span_id); }); ``` `server::InFlightRequest::execute` uses `Context::scope` to set the current context before executing a request, so calls to `context::current` in request handlers will return the context provided by the client. This does not propagate to new spawned tasks. To propagate the client context to child tasks, the following pattern will work: ```rust tokio::spawn(context::current().scope(async { /* do work here */ })); ``` This commit also introduces a breaking change to Context serialization. Previously, the deadline only serialized second-level precision. Now, it provides full fidelity serialization to the nanosecond.	2021-03-13 16:05:02 -08:00