The main() Thread will call tearDown() itself. It appears this
redundancy has existed since e813eaae2f, where the normal error handling
was enhanced at the same time as cleaning up resource management. The
cleanup should have made it obvious the hook was no longer needed, but
alas. Technically, it did originally provide a purpose if setup()
failed, but it would have been better to just move setup() into the
try-catch instead. Today it doesn't even provide that purpose.
The `default_target` field can be unset per the [spec](http://go/grpc-rls-lb-policy-design)
Also fixed a synchronization bug (related to #7460) that `createOrGet()` should be guarded by lock.
AltsChannelBuilder could be improved a bit more by removing the call to
InternalNettyChannelBuilder.setProtocolNegotiatorFactory. However, to do
that cleanest would require reworking how port is plumbed in
NettyChannelBuilder and potentially AbstractManagedChannelImplBuilder to
move getDefaultPort() to ProtocolNegotiator from ClientFactory. Saving
that for another day.
Manually specifying individual Maven artifacts is very verbose and
error-prone. It also does not properly handle transitive dependencies.
It greatly increases the amount of effort to update dependencies.
v1.27.0 introduced support for maven_install and encouraged users to
migrate. I fully expect some users haven't migrated, but it's not clear
that an additional 8 months would help much. Users that don't want to
use maven_install are still free to do so, but would need to maintain
the verbose repository list themselves.
At some point we may begin using the @maven workspace which would
require maven_install, but are not doing so now (except in the examples)
and don't have immediate plans to start.
Add XdsClient implementation of watching LDS/RDS resources, replacing the ConfigWatcher API. This makes LDS/RDS/CDS/EDS resource watchers work similarly. This change also cleans up XdsClientImpl's tests.
Normally the first exception/event experienced is the cause and is
followed by a stampede of ClosedChannelExceptions. In this case,
SslHandler is manufacturing a ClosedChannelException of its own and
propagating it _before_ the trigger event. This might be considered a
bug, but changing SslHandler's behavior would be very risky and almost
certainly break someone's code.
Fixes#7376
Unary are far more common than streaming, and we're throwing for unary even
though it doesn't help the service. Let's stop doing that. We also stop
throwing in onComplete() for all cases, because it doesn't help any service;
it doesn't stop the service's processing and isn't even all that informative
since the cancellation can happen even after onComplete() is called.
`RlsPicker.pickSubchannel()` does not run in SynchronizationContext, but it calls `CachingRlsLbClient.get()` which assumed running in SynchronizationContext. Fixed by removing `synchronizationContext.throwIfNotInThisSynchronizationContext()`. `CachingRlsLbClient.get()` is actually thread-safe in the sense it's guarded by lock, and `DataCacheEntry`'s fields are final.
`ChildPolicyWrapper.picker` was not thread-safe. Fixed by making it volatile.
Changed the test a bit since the old test doesn't really test things well.
It's hoped that this resolves the "too_many_pings" issue some users are
seeing that is worked around by GRPC_EXPERIMENTAL_AUTOFLOWCONTROL=false.
This change also avoids resetting the ping count for empty data frames
(which shouldn't really happen with gRPC).
The previous code failed to reset the ping count on HEADERS and
WINDOW_UPDATE. The code _appeared_ to have callbacks for WINDOW_UPDATE,
but was layered above the Http2Connection so was never called. Thus,
this version is much more aggressive then the previous version while
also addressing the correctness issue.
It deprecates ExpectedException and Assert.assertThat(T, org.hamcrest.Matcher).
Without Java 8 we don't want to migrate away from ExpectedException at
this time. We tend to prefer Truth over Hamcrest, so I swapped the one
instance of Assert.assertThat() to use Truth. With this change we get a
warning-less build with JUnit 4.13. We don't yet upgrade because we
still need to support JUnit 4.12 for some use-cases, but will be able to
upgrade to 4.13 soon when they upgrade.
Introduce ResourceSubscriber for tracking the state of a single resource.
Every time newly subscribing to some resource, a corresponding ResourceSubscriber is created. Note it does not control the resource discovery RPCs. It is still the XdsClient that sends RPCs for with all subscribed resource names for each type. A ResourceSubscriber can have the following states:
- When the initial resource fetch timer (respTimer) is pending, the resource is under discovery, the resource data is unknown. Even if the XdsClient receives a response not containing the corresponding resource, it does not mean the resource is absent. We still need to wait until a response containing the resource data coming or the timer being fired. The timer is scheduled when the ResourceSubscriber is created. So the XdsClient should always create the corresponding ResourceSubscriber when it starts to subscribe a new resource.
- If the resource fetch timer is not pending, we must know the existence of the resource data. If data field is set, it is the most recently received resource data (aka, cached entry). Otherwise, absent field is set to true, indicating the resource does not exist. The exceptional case is when the ADS stream is closed and in the retry backoff period. During that period, respTimer is cancelled and the resource existence may or may not be known. Once the backoff finishes, the XdsClient will reschedule the respTimer when it recreates the ADS stream and re-request all the resources.
Watchers can be added to existing ResourceSubscribers. At the time the watcher is added, its callback will be invoked if we've already known the existence of the resource. Otherwise, the watcher will just sit there and wait data or absence to come in the future.
The PROXYLESS_CLIENT_HOSTNAME node metadata was a temporary workaround for management server to not send back all backend services as load reporting clusters. Now the management server is able to use `send_all_clusters` field to let the client side decide the group of clusters it is reporting loads for. So this node metadata is no longer needed.
Sync Envoy proto to commit fd28e42f31730f5ed6f13f52999692a4885dd312 (2020-09-15, internal Envoy import CL: cl/332279092). Suppressed warning for using listening_addresses in Node and added TODO for cleaning it up.
Updated protobuf gradle plugin version to 0.8.13. Fixed Android Kokoro's memory issue by forcing to use a new Gradle daemon for building the previous commit.
Eric Anderson
ST-DDT
sanjaypujare
Chengyuan Zhang
ZHANG Dapeng
Ran
Sergii Tkachenko