[jira] [Created] (HBASE-18289) TestReplicaWithCluster#testReplicaGetWithPrimaryAndMetaDown is a flaky test

["Mike Drob (JIRA)" <ji...@apache.org>]

Mike Drob created HBASE-18289:
---------------------------------

             Summary: TestReplicaWithCluster#testReplicaGetWithPrimaryAndMetaDown is a flaky test
                 Key: HBASE-18289
                 URL: https://issues.apache.org/jira/browse/HBASE-18289
             Project: HBase
          Issue Type: Test
          Components: test
            Reporter: Mike Drob


Started looking into this today, spend about half the day, and couldn't finish it off, but am filing this JIRA so that I can record progress somewhere and maybe somebody else with more contextual knowledge can chime in.

I'll attach a truncated log file from one of the flaky job runs that focuses on only this test.

We enable the regions are down simulation at 16:31:30 in the file, and we can see that reads on the primary fail and then succeed on the replica for a while. There's a lot of stack traces starting at that point, so I have trouble keeping track of when exactly the replica disappears. Scans of the meta replica look like they work the whole time, it's the user table that fails.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)