Commit Graph

150 Commits

Author SHA1 Message Date
Eric Anderson 2addeae2db NettyClientHandler should handle all exceptionCaught()s
If NettyClientHandler doesn't then the exception will propagate to the
end of the pipeline, get logged, and cause any open calls to hang.
2015-08-10 13:24:03 -07:00
Eric Anderson 849ed1b995 netty: Simplify ping cancellation
ChannelInactive should be called in all cases of channel going down, so
we only need to cancel ping there. Use goAwayStatus for the error, since
we will be putting the most effort into making that status useful.
2015-08-10 13:24:00 -07:00
Eric Anderson 017cdd2ae9 netty: Remove connectionError and just use goAwayStatus
When connectionError was set, goAwayStatus was also set, so we shouldn't
lose any errors.

NettyClientTransport doesn't really need a Throwable, it just needs a
Status. Passing a Status out of NettyClientHandler reduces the number of
places that need to do transport-specific translation of Throwables into
Status codes.
2015-08-10 13:23:27 -07:00
Eric Anderson e1c348c24a netty: Don't double-process ssl handshake failures
As described in SslHandler's documentation, handshakeFuture() and
SslHandshakeCompletionEvent are equivalent forms of learning of
handshake completion. Watching both causes double-logging and serves no
purpose.
2015-08-10 12:38:47 -07:00
Eric Anderson 8d38d03c2f netty: If negotiation fails, continue failing new writes
Otherwise new writes will be written to the channel and will fail in
some unhelpful way.

Logging was removed as we really want to propagate the failure back to
the application via Calls, which is done by failing the
CreateStreamCommand message. Propagating back to the application via
call removes uncontrollable log spam and is necessary anyway to inform
the application what sort of failure occurred in order to appropriately
to react.
2015-08-10 12:38:46 -07:00
Jack Coughlin 3eaa92bea6 Send an RST_STREAM frame on server deadline 2015-08-10 09:51:54 -07:00
Kun Zhang fc85a4085a Add more documentation for transports.
- Add package descriptions for transport, netty and okhttp.
- Describe transports (netty, okhttp and inprocess) in README
2015-08-06 17:24:42 -07:00
Carl Mastrangelo aebb58b200 Change awaitTerminated to awaitTermination 2015-08-06 13:52:17 -07:00
Xudong Ma 7d1e65c111 Switch ALPN/NPN to advertise only h2 2015-08-05 11:25:05 -07:00
Eric Anderson 41d875c7e3 Notify transportReady() in Netty 2015-08-03 16:50:05 -07:00
Eric Anderson 248f575a59 Make any lib-generated UNKNOWN have description
Note that even more importantly, this translates a RST_STREAM error code
to a gRPC status code. This is generally useful, but also necessary for
DEADLINE_EXCEEDED to be more reliable in 0eae0d9.

Fixes #687
2015-08-03 11:15:02 -07:00
Carl Mastrangelo bd8987af1a Add a status to Transport shutdown 2015-07-29 15:35:38 -07:00
Eric Anderson 26d77ecd2e Minor readability changes
Improved some consistency. writeHeaders was the only non-final
implementation method of ServerStream, even though it is really no
different than the others.
2015-07-23 10:57:00 -07:00
Carl Mastrangelo 0003e44886 Add simple server timeout support
Reintroduce throws

Add timeoutExecutor shutdown

Use a default future

Move timeout cancellation

Cancel the timeout in error cases
2015-07-15 09:33:16 -07:00
Jack Coughlin 3e26b993ce Enforce request deadline
Use a ScheduledExecutorService in the ChannelImpl to terminate the
request by closing the ClientStream with status DEADLINE_EXCEEDED
2015-07-14 16:47:45 -07:00
Kun Zhang 73acc73dbf Remove Method and switch its users to MethodDescriptor.
Resolves #511.

- In generated code, make CONFIG private and METHOD_* fields public.
  METHOD_* fields are MethodDescriptors now, users of the CONFIG field
  should switch to using the METHOD_* fields.
- Move MethodType into MethodDescriptor (#529).
- Unify the fully qualified method name. It is fully qualified service
  name + slash + short method name. It doesn't have the leading slash.
- HandlerRegistry switches the key from short method name to fully
  qualified method name.
2015-07-09 09:29:03 -07:00
nmittler a6c51e7e14 Fixing some compiler warnings. 2015-07-08 16:10:54 -07:00
nmittler 8c1d38a0d8 Adding default User-Agent for netty and okhttp. 2015-07-08 15:56:54 -07:00
nmittler efbb65522b Simplifying flow control window config for Netty.
Fixes #494
2015-07-08 15:43:03 -07:00
Kun Zhang d3c5b00827 Add CallOptions.
- Pass CallOptions to Channel.newCall() and
  ClientInterceptor.interceptCall().
- Remove timeout from AbstractStub.StubConfigBuilder and add deadline,
  which is stored in a CallOptions inside the stub.
- Deadline is in nanoseconds in the clock defined by System.nanoTime().
  It is converted to timeout before transmitting on the wire. Fail the
  call with DEADLINE_EXCEEDED if it's already expired.
2015-07-07 14:28:38 -07:00
Eric Anderson c7403127ea Revert swapping to the "canonical HTTP mapping"
The mapping is poorly suited for gRPC. C and Go don't even do any
mapping. We can improve the mapping in the future, but it is very
important that users don't start depending on the current mapping.

This change is "inspired by" the original code, but is even more
conservative.

Fixes #477
2015-06-25 12:15:34 -07:00
nmittler ae2b5bfda7 Exposing AbstractBufferingHandler so it can be used by custom protocol negotiators. 2015-06-24 13:59:33 -07:00
Jack Coughlin 77878a04ee Pass timeout header in ChannelImpl 2015-06-17 16:03:13 -07:00
nmittler 4322a43824 Throw error for Netty when Jetty ALPN not configured. 2015-06-17 14:19:55 -07:00
nmittler 9a6ded523a Cleanup of NettyClientTransportTest. 2015-06-16 13:11:03 -07:00
Eric Anderson 424c7e64c3 Avoid GCM in unit tests
GCM is very slow, and doesn't provide any benefit in unit tests. Even if
we were using tcnative and GCM is fast, using more available ciphers in
tests still makes sense. With this change building with Java 7 works
again, although that isn't the reason for the change.

On my machine with parallel building, it cuts full build time from
92 seconds to 39 seconds. For an incremental build after only changing
an interop test, the build time is cut from 73 seconds to 15 seconds.
2015-06-16 12:36:57 -07:00
nmittler cb486e461d Testing that buffered streams clean up properly upon disconnect. 2015-06-15 13:10:21 -07:00
nmittler 491aad2659 Changing Netty handlers to properly set initialSettings. 2015-06-11 15:06:20 -07:00
Josh Humphries 0d03f89467 add ping 2015-06-10 12:20:08 -04:00
nmittler 4ee2a6584a Allowing Netty TLS bootstrap handler to be sharable.
Fixes #504
2015-06-03 09:42:08 -07:00
Xudong Ma ffcc5f29c0 Remove unused imports.
to make CheckStyle happy.
2015-05-28 16:10:20 -07:00
nmittler 59f6f45cc4 Switching to Netty's ALPN support.
Also adding documentation for using gRPC with TLS-ALPN within Jetty.

Fixes #180
2015-05-28 13:04:10 -07:00
Louis Ryan 641fc288fc Add support for indeterminate length messages. This will make using GRPC easier for non-proto payload types.
Sync to head
2015-05-26 15:44:46 -07:00
Xudong Ma 6affc8dcd5 upgrade okhttp to 2.3 2015-05-22 16:45:13 -07:00
nmittler 65769c26b3 Making client-side negotiation more pluggable. 2015-05-20 14:37:16 -07:00
Eric Anderson 98c6355079 Disable Netty eventloop graceful termination
When shutting down the Netty event loop, we have already guaranteed that
all users of it are no longer running. Doing a shutdownGracefully is
just delaying graceful JVM termination by two seconds. This is very
noticeable for short-lived processes, like our integration tests.

We would actually also prefer to shutdown quickly and get a
RejectedExecutionException for any newly queued tasks, because that
would be a legitimate bug.

shutdown() is deprecated, thus we do shutdownGracefully with a timeout
of 0.
2015-05-15 18:19:57 -07:00
Louis Ryan d54911713a Implement writes to the channel using a dedicated write queue which allows for efficient flush
coalescing
2015-05-15 09:20:13 -07:00
nmittler 8f537e3ec6 updates to test native epoll 2015-05-14 15:33:55 -07:00
nmittler 5efbb6fe73 Allow 100 streams initially rather than 10.
The HTTP/2 spec suggests 100 as the recommended minimum for SETTINGS_MAX_CONCURRENT_STREAMS (https://tools.ietf.org/html/draft-ietf-httpbis-http2-17#section-6.5.2).  We should use this value as our default.
2015-05-13 14:22:46 -07:00
Eric Anderson 1787106cc7 isReady() should return false until stream allocated
isReady() can provide pushback while the call is in progress, but it
can also provide the pushback necessary when the client creates more
streams than permitted by MAX_CONCURRENT_STREAMS.

As part of this commit, OkHttp is now calling onReady() after call
creation (it previously never called onReady()).
2015-05-12 16:41:50 -07:00
nmittler c3fa600eae Fixing compiler warning. 2015-05-12 16:23:39 -07:00
nmittler d5727c7fcd Deferring stream creation until receiving SETTINGS from server.
Additionally:
- Fixed bug where the decoder was given the incorrect encoder.
- Adding proper logging class for client/server.
2015-05-12 15:46:32 -07:00
nmittler b6407c4a10 Only complete graceful shutdown after buffered streams complete.
Upgrading to to the latest Netty as well.
2015-05-07 08:36:22 -07:00
nmittler b5443833d9 Fixing build issue. 2015-05-07 07:50:33 -07:00
nmittler d8d8ebd783 Adding logging to NettyClientHandler. 2015-05-06 14:38:22 -07:00
nmittler 64df428e36 Properly removing buffered streams after receiving goAway 2015-05-06 14:24:29 -07:00
nmittler 7f73189e56 Always checking MAX_CONCURRENT_STREAMS in
BufferingHttp2ConnectionEncoder.

Currently we don't check this setting when handling a streamClosed()
event. If the setting has lowered prior to this event, the stream
creation could fail.
2015-05-05 19:29:34 -07:00
nmittler a45e0a4767 Cleaning up closing for Netty client/server. 2015-05-05 15:43:22 -07:00
nmittler ef6a2f02d4 Updating to the latest Netty version.
Also re-enabling the gracefulShutdown test now that Netty has been
fixed.
2015-05-05 15:39:21 -07:00
Eric Anderson 10fb20650d Produce combined JavaDoc, add links, exclude internals 2015-05-05 11:32:15 -07:00