[extension/jaegarremotesamplingextension/internal] Leaking goroutine on shutdown

[crobert-1]

Component(s)

extension/jaegerremotesampling

Describe the issue you're reporting

When attempting to add goleak to the internal package of the jaegarremotesampling extension, the following test fails:

=== RUN   TestSamplingGRPCServer_Shutdown
--- PASS: TestSamplingGRPCServer_Shutdown (10.00s)
=== RUN   TestSamplingGRPCServer_Shutdown/graceful_stop_is_successful_without_delay
    --- PASS: TestSamplingGRPCServer_Shutdown/graceful_stop_is_successful_without_delay (0.00s)
=== RUN   TestSamplingGRPCServer_Shutdown/graceful_stop_is_successful_with_delay
    --- PASS: TestSamplingGRPCServer_Shutdown/graceful_stop_is_successful_with_delay (5.00s)
=== RUN   TestSamplingGRPCServer_Shutdown/context_timed_out
    --- PASS: TestSamplingGRPCServer_Shutdown/context_timed_out (5.00s)
=== RUN   TestSamplingGRPCServer_Shutdown/grpc_server_not_started
    --- PASS: TestSamplingGRPCServer_Shutdown/grpc_server_not_started (0.00s)
PASS
goleak: Errors on successful test run: found unexpected goroutines:
[Goroutine 35 in state sleep, with time.Sleep on top of the stack:
time.Sleep(0xdf8475800)
	/usr/local/go/src/runtime/time.go:195 +0x125
github.com/open-telemetry/opentelemetry-collector-contrib/extension/jaegerremotesampling/internal.(*grpcServerMock).GracefulStop(0x0?)
	/Users/crobert/dev/opentelemetry-collector-contrib/extension/jaegerremotesampling/internal/grpc_test.go:91 +0x16
github.com/open-telemetry/opentelemetry-collector-contrib/extension/jaegerremotesampling/internal.(*SamplingGRPCServer).Shutdown.func1()
	/Users/crobert/dev/opentelemetry-collector-contrib/extension/jaegerremotesampling/internal/grpc.go:95 +0x2f
created by github.com/open-telemetry/opentelemetry-collector-contrib/extension/jaegerremotesampling/internal.(*SamplingGRPCServer).Shutdown in goroutine 34
	/Users/crobert/dev/opentelemetry-collector-contrib/extension/jaegerremotesampling/internal/grpc.go:94 +0xa5
]

The actual failing test is TestSamplingGRPCServer_Shutdown/context_timed_out.

The leaked goroutine is started here:

opentelemetry-collector-contrib/extension/jaegerremotesampling/internal/grpc.go

Line 94 in 8ccadc3

go func() {

What's happening is the Shutdown method for the sampling GRPC server starts a goroutine that attempts to gracefully shutdown the server. As soon as the context is cancelled though (graceful stop took too long), Stop is called and the main routine will return, leaking the sub-goroutine that's attempting GracefulStop.

[github-actions]

Pinging code owners:

• extension/jaegerremotesampling: @yurishkuro @frzifus

See Adding Labels via Comments if you do not have permissions to add labels yourself.

[yurishkuro]

@crobert-1 are you addressing this yourself or looking for help?

[crobert-1]

Looking for help at the moment. If no one's able to take this soon I'll plan on eventually proposing a solution. 👍

[chirag-ghosh]

I would like to take this up