[IMPALA-10592] Exhaustive tests timeout after 20 hours

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: Impala 4.0.0
• Fix Version/s: Impala 4.0.0
• Epic Color: ghx-label-9
• Assignee: Andrew Sherman
• Reporter: Andrew Sherman
• Created: 18/Mar/21 00:42
• Updated: 23/Mar/21 16:52
• Resolved: 23/Mar/21 16:52

Description

The tests seem to make progress for nearly 10 hours, but after 20 hours they timeout

**** run-all-tests.sh TIMED OUT! ****

The timeout stack traces are attached

Impala logs show a long period of inactivity between 03/16 16:58 and 03/17 04:53
For example:

I0316 16:56:33.555305  9911 impala-server.cc:1996] Catalog topic update applied with version: 65701 new min catalog object version: 36078
I0316 16:58:00.504211  9041 krpc-data-stream-mgr.cc:427] Reduced stream ID cache from 6 items, to 5, eviction took: 0
I0316 16:58:10.504297  9041 krpc-data-stream-mgr.cc:427] Reduced stream ID cache from 5 items, to 4, eviction took: 0
I0316 16:58:20.504348  9041 krpc-data-stream-mgr.cc:427] Reduced stream ID cache from 4 items, to 3, eviction took: 0
I0316 16:58:30.504386  9041 krpc-data-stream-mgr.cc:427] Reduced stream ID cache from 3 items, to 2, eviction took: 0
I0316 16:58:40.504467  9041 krpc-data-stream-mgr.cc:427] Reduced stream ID cache from 2 items, to 1, eviction took: 0
I0316 16:58:50.504545  9041 krpc-data-stream-mgr.cc:427] Reduced stream ID cache from 1 items, to 0, eviction took: 0
I0317 04:53:06.368000  9905 TAcceptQueueServer.cpp:340] New connection to server StatestoreSubscriber from client <Host: ::ffff:127.0.0.1 Port: 32818>
I0317 04:53:06.368041  9910 thrift-util.cc:96] TSocket::write_partial() send() <Host: ::ffff:127.0.0.1 Port: 36850>Broken pipe
W0317 04:53:06.369092  8780 init.cc:214] A process pause was detected for approximately 18s920ms
I0317 04:53:06.369904  9905 TAcceptQueueServer.cpp:340] New connection to server StatestoreSubscriber from client <Host: ::ffff:127.0.0.1 Port: 32822>
I0317 04:53:06.369961  9910 thrift-util.cc:96] TAcceptQueueServer client died: write() send(): Broken pipe
W0317 04:53:06.369966  8929 JvmPauseMonitor.java:205] Detected pause in JVM or host machine (eg GC): pause of approximately 18338ms
No GCs detected
I0317 04:53:06.370081 27248 thrift-util.cc:96] TSocket::write_partial() send() <Host: ::ffff:127.0.0.1 Port: 32818>Broken pipe
I0317 04:53:06.370126 27248 thrift-util.cc:96] TAcceptQueueServer client died: write() send(): Broken pipe

Issue Links

• is caused by: IMPALA-10596 TestAdmissionControllerStressWithACService.test_mem_limit fails with "Invalid or unknown query handle" when canceling a query