Adar Dembo created KUDU-1669:
--------------------------------
Summary: Java ITClient test can orphan processes
Key: KUDU-1669
URL: https://issues.apache.org/jira/browse/KUDU-1669
Project: Kudu
Issue Type: Bug
Components: client
Affects Versions: 1.0.0
Reporter: Adar Dembo
Saw this with ITClient but it can happen to any test that uses a MiniCluster,
restarts a process, and interrupts its own threads. Maybe ITClient is the only
example of that.
Here's the interesting stack trace:
{noformat}
21:18:12.316 [ERROR - Thread-8] (ITClient.java:134) Couldn't restart a master
java.lang.InterruptedException: sleep interrupted
at java.lang.Thread.sleep(Native Method)
at
org.apache.kudu.client.MiniKuduCluster.configureAndStartProcess(MiniKuduCluster.java:239)
at
org.apache.kudu.client.MiniKuduCluster.restartDeadProcessOnPort(MiniKuduCluster.java:282)
at
org.apache.kudu.client.MiniKuduCluster.restartDeadMasterOnPort(MiniKuduCluster.java:256)
at
org.apache.kudu.client.BaseKuduTest.restartLeaderMaster(BaseKuduTest.java:431)
at
org.apache.kudu.client.ITClient$ChaosThread.restartMaster(ITClient.java:222)
at org.apache.kudu.client.ITClient$ChaosThread.run(ITClient.java:161)
at java.lang.Thread.run(Thread.java:745)
{noformat}
Because of this thread interruption, the newly restarted process never makes it
into BaseKuduTest.masterProcesses, which means it isn't destroyed when the test
class cleans up. All sorts of bad stuff can happen then. It's possible the
process is completely orphaned on the test machine (though I imagine we'd kill
it eventually). I noticed this because in one of my precommit test runs
ITClient left behind a test directory; presumably because the orphaned master
continued to write files even after the test directory was cleaned up.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
