[
https://issues.apache.org/jira/browse/HBASE-18290?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16067356#comment-16067356
]
Mike Drob commented on HBASE-18290:
-----------------------------------
Do you think this is also the root behind TestRegionObserverInterface
flakiness? All the errors there are during
{noformat}
org.junit.runners.model.TestTimedOutException: test timed out after 300000
milliseconds
at java.lang.Thread.sleep(Native Method)
at
org.apache.hadoop.hbase.client.HBaseAdmin$ProcedureFuture.waitProcedureResult(HBaseAdmin.java:3409)
at
org.apache.hadoop.hbase.client.HBaseAdmin$ProcedureFuture.get(HBaseAdmin.java:3339)
at org.apache.hadoop.hbase.client.HBaseAdmin.get(HBaseAdmin.java:1962)
at
org.apache.hadoop.hbase.client.HBaseAdmin.deleteTable(HBaseAdmin.java:558)
at
org.apache.hadoop.hbase.HBaseTestingUtility.deleteTable(HBaseTestingUtility.java:1747)
at
org.apache.hadoop.hbase.coprocessor.TestRegionObserverInterface.testRecovery(TestRegionObserverInterface.java:616)
{noformat}
or because it is a timeout it might be a different failure?
> Fix TestAddColumnFamilyProcedure and TestDeleteTableProcedure
> -------------------------------------------------------------
>
> Key: HBASE-18290
> URL: https://issues.apache.org/jira/browse/HBASE-18290
> Project: HBase
> Issue Type: Bug
> Components: test
> Affects Versions: 2.0.0
> Reporter: stack
> Assignee: stack
> Fix For: 2.0.0
>
>
> These two tests don't pass. Turns out the cause was interesting.
> We added a workaround for case where procedure WAL could have procs out of
> order.
> HBASE-18216 [AMv2] Workaround for HBASE-18152, corrupt procedure WAL
> If we find a procedure that is not 'increasing' -- of a later timestamp or
> procid -- then we'd skip the application of the 'old' proc. The workaround
> was until we figure in what scenarios we can write procedures out of order
> (seems to be rare and high-concurrency... TBD).
> These two tests trip FAILs and ROLLBACKs (double delete of table or disable
> of an already disabled table). They are good tests. But procedures that get
> marked FAIL or ROLLEDBACK will have procids that are less than current. Makes
> it so we skipped adding the ROLLBACK and so finishing up the procedure.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
