Java performance benchmarks dirty shutdown

[jtattermusch]

https://grpc-testing.appspot.com/view/Performance/job/gRPC_performance_experiment/4231/consoleFull

In most of scenarios, the java qps_workers don't seem to shutdown cleanly:

2017-06-26 12:32:42,813 WAITING: ETA 66.3 sec; 0 queued, 1 jobs running, 1 complete, 0 failed
2017-06-26 12:32:42,860 E0626 12:32:42.832971879   11903 driver.cc:501]              Worker 0 could not be properly quit because Received RST_STREAM with error code 8
E0626 12:32:42.847037641   11903 driver.cc:501]              Worker 1 could not be properly quit because Received RST_STREAM with error code 8
E0626 12:32:42.859240520   11903 driver.cc:501]              Worker 2 could not be properly quit because Received RST_STREAM with error code 8

2017-06-26 12:32:42,860 FAILED: qps_json_driver.quit [ret=1, pid=11902]

The log also has occasional [GC (Allocation Failure)] messages:

[GC (Allocation Failure) [PSYoungGen: 524800K->11962K(611840K)] 524800K->11978K(2010112K), 0.0112577 secs] [Times: user=0.04 sys=0.00, real=0.02 secs] 
[GC (Allocation Failure) [PSYoungGen: 536762K->4660K(611840K)] 536778K->4684K(2010112K), 0.0062845 secs] [Times: user=0.03 sys=0.01, real=0.01 secs] 
[GC (Allocation Failure) [PSYoungGen: 529460K->4580K(611840K)] 529484K->4612K(2010112K), 0.0056438 secs] [Times: user=0.02 sys=0.00, real=0.00 secs] 

[zpencer]

I can take a look into this as I am updating our code to include the missing server and client workloads.

[zpencer]

@jtattermusch is this one still happening?

[jtattermusch]

I am not sure, can you please inspect the test logs yourself? Recent job runs can be found here: https://kokoro2.corp.google.com/job/grpc/job/core/job/master/job/linux/job/grpc_full_performance_master/

[zpencer]

Ah thanks for the updated link. The RST_STREAM errors are still happening. The GC allocation failures can be ignored however, they just indicate a garbage collection took place.