First take for grpclb selection stabilization:
1. Changed DnsNameResolver to return balancer addresses as a GrpcAttributes.ATTR_LB_ADDRS attribute in ResolutionResult, instead of among the addresses.
2. AutoConfiguredLoadBalancerFactory decides LB policy solely based on parsed service config without looking at resolved addresses. Behavior changes:
- If no LB policy is specified in service config, default to pick_first, even if there exist balancer addresses (in attributes).
- If grpclb specified but not available and no other specified policies available, it will fail without fallback to round_robin.
3. GrpclbLoadBalancer populates balancer addresses from ResolvedAddresses's attribute (GrpclbConstants.ATTR_LB_ADDRS) instead of sieving from addresses.
Go to TRANSIENT_FAILURE immediately instead of NOOP if GracefulSwitchLoadBalancer receives resolution error before switching to any delegate.
In most natural usecase, the `gracefulSwitchLb` is a child balancer of some `parentLb`, and the `gracefulSwitchLb` switches to a new `delegateLb` when `parentLb.handleResolvedAddressGroup()`. If the `parentLb` receives a resolution error before receiving any resolved addresses, it should go to TRANSIENT_FAILURE. In this case, it will be more convenient if the initial `gracefulSwitchLb` can go to TRANSIENT_FAILURE directly.
Decouples grpc-core with census, while still preserve the default integration of census in grpc-core. Users wishing to enable census needs to add grpc-census to their runtime classpath.
- Created a grpc-census module:
- Moved CensusStatsModule.java and CensusTracingModule.java into grpc-census from grpc-core. CensusModuleTests.java is also moved. They now belong to io.grpc.census package.
Moved DeprecatedCensusConstants.java into io.grpc.census.internal (is this necessary?) in grpc-census.
- Created CensusStatsAccessor.java and CensusTracingAccessor.java, which are used to create census ClientInterceptor and ServerStreamTracer.Factory.
- Everything in grpc-census are package private, except the accessor classes. They only publicly expose ClientInterceptor and ServerStreamTracer.Factory, no Census specific types are exposed.
- Use runtime reflection to load and apply census stats/tracing to channel/server builders, if grpc-census is found in runtime classpath.
- Removed special APIs on AbstractManagedChannelImplBuilder and AbstractServerImplBuilder for overriding census module. They are only used for testing. Now we changed tests to apply Census ClientInterceptor and ServerStreamTracer.Factory just as normal interceptor/stream tracer factory. Test writer is responsible for taking care of the ordering concerns of interceptors and stream tracer factories.
This PR is to add one more Executor parameter when creating the SslContext.
In Netty, we already have this implementation for passing Executor when creating SslContext: netty/netty#8847. This extra Executor is used to take some time-consuming tasks when doing SSL handshake. However, in current gRPC implementation, we are not using this API.
In this PR, the relevant changes are:
1. get the executorPool from ChannelBuilder or ServerBuilder
2. pass the executorPool all the way down to ClientTlsHandler
3. fill executorPool in when creating SslHandler
`GracefulSwitchLoadBalancer` was doing switch based on `LoadBalancerProvider.getPolicyName()`. This turned out to be very awkward when I have to synthesize a policy name for the provider, and what I actually care about is the identity of the lb provider not necessarily the policy name.
Now `GracefulSwitchLoadBalancer` is doing switch based on identity of `LoadBalancer.Factory`, which is simpler.
The API review for #6279 came up with a more meaningful name that
better explains the intent. A setter for the old name was left in
ManagedChannelBuilder to ease migration to the new name by current
users.
Adds an Executor to NameResolver.Args, which is optionally set on ManagedChannelBuilder. This allows NameResolver implementations to avoid creating their own thread pools if the application already manages its own pools.
Addresses #3703.
This allows plumbing information to the subchannel via Attributes, like
an authority override. RR still does not support multiple EAGs that only
differ by attributes.
This can be used to prevent duplicate classes in the classpath, one via
Maven and one via Bazel-native.
See census-instrumentation/opencensus-java#1963 and #5359
ServerImpl uses that ticker to create incoming Deadlines. This feature is specifically restricted to in-process, as it can also customize ScheduledExecutorService, and them together can fake out the clock which is useful in tests. On the other hand, a fake Ticker won't work with Netty's ScheduledExecutorService.
Also improved mismatch detection, documentation and tests in Deadline.
Fixes#5593 and supersedes #5601
Now that https://github.com/census-instrumentation/opencensus-java/pull/1854 has been merged & released as 0.21.0. We can start using the method & status tags.
Background:
Opencensus introduced new tags for status and method (https://github.com/census-instrumentation/opencensus-java/pull/1115). The old views that used those tags were deprecated and new views were created that used the new tags. However grpc-java wasn't updated to use the new tags due to concern of breaking existing metrics. This resulted in the old views being deprecated while the new views were broken (goomics #50).
https://github.com/census-instrumentation/opencensus-java/pull/1854 added a compatibility layer to opencensus that would remap new tags to old tags for old views. This should unblock grpc to switching to the new tags while allowing old views to still be populated. That commit was released as part of opencensus 0.21, which grpc currently uses
* Bump PerfMark to 0.17.0
The main changes how linking is done. Linking is now always done
through the `PerfMark` entry class. This is for two reasons:
1. It make instrumenting the linking calls *much* easier.
2. It follows the API pattern of "verbNoun()". Previous callsites
would have `Link link = PerfMark.link(); link.link()`. This
stuttering is not quick to follow.
Generated using:
```
find -name \*.java -exec sed -i 's#link = PerfMark.link();#link = PerfMark.linkOut();#g' {} \;
find -name \*.java -exec sed -i 's#link.link();#PerfMark.linkIn(link);#g' {} \;
find -name \*.java -exec sed -i 's#command.getLink().link();#PerfMark.linkIn(command.getLink());#g' {} \;
find -name \*.java -exec sed -i 's#cmd.getLink().link();#PerfMark.linkIn(cmd.getLink());#g' {} \;
find -name \*.java -exec sed -i 's#msg.getLink().link();#PerfMark.linkIn(msg.getLink());#g' {} \;
```
Since the deprecated link methods are also `@DoNotCall`, the same
sed calls will need to be used on import.
Works for #4740
- Subclasses of `AbstractClientStream` include remote address in insight if available.
- `DelayedStream` adds buffered time, and the insight of real stream if it's set.
- `RetriableStream` insights outputs of Substreams.
Example error message:
```
deadline exceeded after 8112071ns. [buffered_nanos=24763, remote_addr=foo.test.google.fr/127.0.0.1:44749]
```
or
```
deadline exceeded after 8112071ns. [buffered_nanos=22344324763, waiting_for_connection]
```
This is related to #4776 but taking a more usage-specific approach.
In #5892 getAttributes() is called without any regard of timing.
Currently DelayedStream.getAttributes() wil throw if called before
passThrough was set. Just to be safe, we are removing that
restriction and making it clear on the javadoc.
On the other hand, we intend to keep the timing restriction on
ClientCall.getAttributes().
This change removes the WriteQueue linking and splits it out into each
of the commands, so that the trace is more precise, and the tag
information is correct.
It is still unclear what the initial Tag should be for ClientCallImpl,
since it should not access the TransportState to get the HTTP/2 stream id.
No more methods on the `LoadBalancer` will be called after
`LoadBalancer#shutdown()` is called. This includes
`LoadBalancer#handleSubchannelState()` too. `SubchannelStateListener`
inherited this restriction. However, this special case makes
`onSubchannelState(SHUTDOWN)` an unreliable way of being notified
about `Subchannel` SHUTDOWN, and may confuse/complicate a
wrapping `LoadBalancer` that expects the full notification (e.g., #5875).
The javadoc isn't clear whether this restriction applies. I think
it's more useful to make it no apply.
We were logging when withDeadline() was used, not when the Context was used. As
discovered while looking at https://stackoverflow.com/q/56593692/4690866 .
In e19e8f7d updateTimeoutHeaders was removed and logIfContextNarrowedTimeout
was called directly. However, the two methods had reverse ordering of
callDeadine/outerCallDeadline and the caller did not get their arguments
swapped.
panic mode was temporarily disabled by #4152 and re-enabled by #4245,
but the tests were not. This has caused a few test code that was
broken but not executed at all.
This add perfmark annotations in some key places, notably on transport/application boundaries, and thread hop locations. Perfmark records to a thread-local buffer the events that happen in each thread. Perfmark is disabled by default, and will compile to a noop unless Perfmark.setEnabled is invoked. This should make it free when disable, and pretty fast when it is enabled.
It is important that started tasks are ended, so several places in our code are moved to either try-finally blocks, or moved into a private method. I realize this is ugly, but I think it is manageable. In the future, we can look at making an agent or compiler plugin that simplifies the recording.
Linking between threads is done with a Link object, which is created on the "outbound" task, and used on the "inbound" task. This is slightly more verbose, and does has a small amount of runtime overhead, even when disabled. (for null checks, slightly higher memory usage, etc.) I think this is okay to, because it makes other optimizations much easier.
Previously, AutoConfiguredLoadBalancer was also handling it, but it
doesn't trigger retries. By returning true for
canHandleEmptyAddressListFromNameResolution(),
AutoConfiguredLoadBalancer effectively by-passed the empty-result
handling logic in ManagedChannelImpl, thus resolution retries were
never triggered.
This change requires AutoConfiguredLoadBalancer to stop being a
LoadBalancer, for its tryHandleResolvedAddresses(). It doesn't
cause any trouble because AutoConfiguredLoadBalancer has become
less and less like a LoadBalancer during the service config changes.
As mentioned in 5188[1], the default used with the enableKeepAlive API
conflicted with the default server enforcement. Instead of fixing it,
remove it. These APIs were deprecated in v1.3.0 in April 2017.
1. https://github.com/grpc/grpc-java/issues/5188#issuecomment-482269303
The `method` and `status` shouldn't be propagated in the first place,
but in previous OpenCensus implementation all tags are propagating by
default. Now with the TagMetadata it may make sense to change them to
local tags.
This will be a breaking change to users who depend on the behavior that
these tags propagate through process boundaries.
The old createSubchannel() doesn't require being called from
sync-context, thus its implementation schedules start() on
sync-context because start() has such requirement. However, if a
LoadBalancer call createSubchannel() in sync-context, start() is
queued and only called after the control exits the sync-context. If
createSubchannel() is immediately followed by other Subchannel methods
that requires start(), such as requestConnection(), they will fail.
This will be a very common issue as most LoadBalancers call
createSubchannel() in the sync-context.
This fix splits out the real work of start() into internalStart() which
is called by the old createSubchannel().
Jihun Cho
ZHANG Dapeng
Chengyuan Zhang
Eric Gribkoff
Eric Anderson
Tomo Suzuki
Graeme Morgan
ZhenLian
Ran
Yongwoo Noh
Carl Mastrangelo
Kun Zhang
Grant Oakley
edr
Aaditya Gavandalkar
Dharmesh Jogadia
Igor Bernstein
Carl Mastrangelo
Yang Song