[
https://issues.apache.org/jira/browse/HBASE-20642?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16491348#comment-16491348
]
Ankit Singhal commented on HBASE-20642:
---------------------------------------
bq. The procedure was submitted, right, and started to make progress (it got as
far as changing the table descriptor?). Did the procedure not succeed? Though
there was a crash of Master in the middle of its running? If it did not
complete, that is a problem.
bq. Sounds like the original procedure did not complete? Is that so? That it
died in the middle of its running and so you tried to resubmit the add
column... but it fails because the original procedure died half-way through? Is
this what is happening?
No, The procedure will get succeed eventually after replaying procedure WALs,
it's just user will get InvalidFamilyOperationException even for the first
attempt only.
bq.You mean, a user will retry because they think their original submission did
not take? In this case, if a Procedure in-flight modifying the table, this
second submission should fail.
It's actually not the user, user is making a call only once but HBase client
itself retries the call while master is restarting and if master come back in
between and the procedure is completed, user will see
InvalidFamilyOperationException because HBase consider it as a second call from
the user although it is coming as part of retry by HBase client.
So the patch is to move all the checks in Procedure so that we do nonce check
to differentiate whether it is a retry or new call before actually executing
them.
> IntegrationTestDDLMasterFailover throws 'InvalidFamilyOperationException
> -------------------------------------------------------------------------
>
> Key: HBASE-20642
> URL: https://issues.apache.org/jira/browse/HBASE-20642
> Project: HBase
> Issue Type: Bug
> Reporter: Ankit Singhal
> Assignee: Ankit Singhal
> Priority: Major
> Attachments: HBASE-20642.patch
>
>
> [~romil.choksi] reported that IntegrationTestDDLMasterFailover is failing
> while adding column family during the time master is restarting.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
