tarpc

Prepare release of tarpc v0.35.0

Prepare for v0.34.0 release

Breaking Changes

Opentelemetry dependency version increased to 0.18.

Breaking Changes

• As part of a fix to return more channel errors in RPC results, a few error types have changed:

  1. client::RpcError::Disconnected was split into the following errors:
  • Shutdown: the client was shutdown, either intentionally or due to an error. If due to an error, pending RPCs should see the more specific errors below.
  • Send: an RPC message failed to send over the transport. Only the RPC that failed to be sent will see this error.
  • Receive: a fatal error occurred while receiving from the transport. All in-flight RPCs will receive this error.
  1. client::ChannelError and server::ChannelError are unified in tarpc::ChannelError. Previously, server transport errors would not indicate during which activity the transport error occurred. Now, just like the client already was, it will be specific: reading, readying, sending, flushing, or closing.

New Features

This release adds Unix Domain Sockets to the serde_transport module. To use it, enable the "unix" feature. See the docs for more information.

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.

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.

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.

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.

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.

This version was yanked due to tarpc-plugins version mismatches.

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

Breaking Changes

Major server module refactoring

1. Renames

Some of the items in this module were renamed to be less generic:

• Handler => Incoming
• ClientHandler => Requests
• ResponseHandler => InFlightRequest
• Channel::{respond_with => requests}

In the case of Handler: handler of what? Now it's a bit clearer that this is a stream of Channels (aka incoming connections).

Similarly, ClientHandler was a stream of requests over a single connection. Hopefully Requests better reflects that.

ResponseHandler was renamed InFlightRequest because it no longer contains the serving function. Instead, it is just the request, plus the response channel and an abort hook. As a result of this, Channel::respond_with underwent a big change: it used to take the serving function and return a ClientHandler; now it has been renamed Channel::requests and does not take any args.

1. Execute methods

All methods thats actually result in responses being generated have been consolidated into methods named execute:

• InFlightRequest::execute returns a future that completes when a response has been generated and sent to the server Channel.
• Requests::execute automatically spawns response handlers for all requests over a single channel.
• Channel::execute is a convenience for channel.requests().execute().
• Incoming::execute automatically spawns response handlers for all requests over all channels.

1. Removal of Server.

server::Server was removed, as it provided no value over the Incoming/Channel abstractions. Additionally, server::new was removed, since it just returned a Server.

Client RPC methods now take &self

This required the breaking change of removing the Client trait. The intent of the Client trait was to facilitate the decorator pattern by allowing users to create their own Clients that added behavior on top of the base client. Unfortunately, this trait had become a maintenance burden, consistently causing issues with lifetimes and the lack of generic associated types. Specifically, it meant that Client impls could not use async fns, which is no longer tenable today, with channel libraries moving to async fns.

Servers no longer send deadline-exceed responses.

The deadline-exceeded response was largely redundant, because the client shouldn't normally be waiting for such a response, anyway -- the normal client will automatically remove the in-flight request when it reaches the deadline.

This also allows for internalizing the expiration+cleanup logic entirely within BaseChannel, without having it leak into the Channel trait and requiring action taken by the Requests struct.

Clients no longer send cancel messages when the request deadline is exceeded.

The server already knows when the request deadline was exceeded, so the client didn't need to inform it.

Fixes

• When a channel is dropped, all in-flight requests for that channel are now aborted.

Breaking Changes

Upgrades tokio to 1.0.

This release was yanked.

Breaking Changes

Upgrades tokio to 0.3.

This release adds some flexibility and consistency to serde_transport, with one new feature and one small breaking change.

New Features

serde_transport::tcp now exposes framing configuration on connect() and listen(). This is useful if, for instance, you want to send requests or responses that are larger than the maximum payload allowed by default:

let mut transport = tarpc::serde_transport::tcp::connect(server_addr, Json::default);
transport.config_mut().max_frame_length(4294967296);
let mut client = MyClient::new(client::Config::default(), transport.await?).spawn()?;

Breaking Changes

The codec argument to serde_transport::tcp::connect changed from a Codec to impl Fn() -> Codec, to be consistent with serde_transport::tcp::listen. While only one Codec is needed, more than one person has been tripped up by the inconsistency between connect and listen. Unfortunately, the compiler errors are not much help in this case, so it was decided to simply do the more intuitive thing so that the compiler doesn't need to step in in the first place.

New Features

#[tarpc::server] diagnostics

When a service impl uses #[tarpc::server], only async fns are re-written. This can lead to confusing compiler errors about missing associated types:

error: not all trait items implemented, missing: `HelloFut`
 --> $DIR/tarpc_server_missing_async.rs:9:1
  |
9 | impl World for HelloServer {
  | ^^^^

The proc macro now provides better diagnostics for this case:

error: not all trait items implemented, missing: `HelloFut`
 --> $DIR/tarpc_server_missing_async.rs:9:1
  |
9 | impl World for HelloServer {
  | ^^^^

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 {
   |     ^^

Bug Fixes

Fixed client hanging when server shuts down

Previously, clients would ignore when the read half of the transport was closed, continuing to write requests. This didn't make much sense, because without the ability to receive responses, clients have no way to know if requests were actually processed by the server. It basically just led to clients that would hang for a few seconds before shutting down. This has now been corrected: clients will immediately shut down when the read-half of the transport is closed.

More docs.rs documentation

Previously, docs.rs only documented items enabled by default, notably leaving out documentation for tokio and serde features. This has now been corrected: docs.rs should have documentation for all optional features.

New Features

A new proc macro, #[tarpc::server] was added! This enables service impls to elide the boilerplate of specifying associated types for each RPC. With the ubiquity of async-await, most code won't have nameable futures and will just be boxing the return type anyway. This macro does that for you.

Breaking Changes

• Enums had _non_exhaustive fields replaced with the #[non_exhaustive] attribute.

Bug Fixes

• https://github.com/google/tarpc/issues/304

  A race condition in code that limits number of connections per client caused occasional panics.

• https://github.com/google/tarpc/pull/295

  Made request timeouts account for time spent in the outbound buffer. Previously, a large outbound queue would lead to requests not timing out correctly.

Breaking Changes

1. tarpc has updated its tokio dependency to the latest 0.2 version.
2. The tarpc crates have been unified into just tarpc, with new Cargo features to enable functionality.
   • The bincode-transport and json-transport crates are deprecated and superseded by the serde_transport module, which unifies much of the logic present in both crates.

Breaking Changes

Version 0.13 marks a significant departure from previous versions of tarpc. The API has changed significantly. The tokio-proto crate has been torn out and replaced with a homegrown rpc framework. Additionally, the crate has been modularized, so that the tarpc crate itself contains only the macro code.

New Crates

• crate rpc contains the core client/server request-response framework, as well as a transport trait.
• crate bincode-transport implements a transport that works almost exactly as tarpc works today (not to say it's wire-compatible).
• crate trace has some foundational types for tracing. This isn't really fleshed out yet, but it's useful for in-process log tracing, at least.

All crates are now at the top level. e.g. tarpc-plugins is now tarpc/plugins rather than tarpc/src/plugins. tarpc itself is now a very small code surface, as most functionality has been moved into the other more granular crates.

New Features

• deadlines: all requests specify a deadline, and a server will stop processing a response when past its deadline.
• client cancellation propagation: when a client drops a request, the client sends a message to the server informing it to cancel its response. This means cancellations can propagate across multiple server hops.
• trace context stuff as mentioned above
• more server configuration for total connection limits, per-connection request limits, etc.

Removals

• no more shutdown handle. I left it out for now because of time and not being sure what the right solution is.
• all async now, no blocking stub or server interface. This helps with maintainability, and async/await makes async code much more usable. The service trait is thusly renamed Service, and the client is renamed Client.
• no built-in transport. Tarpc is now transport agnostic (see bincode-transport for transitioning existing uses).
• going along with the previous bullet, no preferred transport means no TLS support at this time. We could make a tls transport or make bincode-transport compatible with TLS.
• a lot of examples were removed because I couldn't keep up with maintaining all of them. Hopefully the ones I kept are still illustrative.
• no more plugins!

Breaking Changes

Updates tarpc to use tarpc-plugins 0.2.

Breaking Changes

This release updates tarpc to use serde 1.0. As such, users must also update to use serde 1.0. The serde 1.0 release notes detail migration paths.

This release removes the Sync bound on RPC args for both sync and future clients. No breaking changes.

Breaking Changes

This release updates tarpc-plugins to work with rustc master. Thus, older versions of rustc are no longer supported. We chose a minor version bump because it is still source-compatible with existing code using tarpc.

This release was purely doc fixes. No breaking changes.