[ https://issues.apache.org/jira/browse/HDFS-2738?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13179315#comment-13179315
]
Hari Mankude commented on HDFS-2738:
------------------------------------
This seems to be easily reproducible by doing couple of manual failover round-trips between
namenodes when write traffic is happening.
> FSEditLog.selectinputStreams is reading through in-progress streams even when non-in-progress
are requested
> -----------------------------------------------------------------------------------------------------------
>
> Key: HDFS-2738
> URL: https://issues.apache.org/jira/browse/HDFS-2738
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: ha, name-node
> Affects Versions: HA branch (HDFS-1623)
> Reporter: Todd Lipcon
> Assignee: Aaron T. Myers
> Priority: Blocker
>
> The new code in HDFS-1580 is causing an issue with selectInputStreams in the HA context.
When the active is writing to the shared edits, selectInputStreams is called on the standby.
This ends up calling {{journalSet.getInputStream}} but doesn't pass the {{inProgressOk=false}}
flag. So, {{getInputStream}} ends up reading and validating the in-progress stream unnecessarily.
Since the validation results are no longer properly cached, {{findMaxTransaction}} also re-validates
the in-progress stream, and then breaks the corruption check in this code. The end result
is a lot of errors like:
> 2011-12-30 16:45:02,521 ERROR namenode.FileJournalManager (FileJournalManager.java:getNumberOfTransactions(266))
- Gap in transactions, max txnid is 579, 0 txns from 578
> 2011-12-30 16:45:02,521 INFO ha.EditLogTailer (EditLogTailer.java:run(163)) - Got error,
will try again.
> java.io.IOException: No non-corrupt logs for txid 578
> at org.apache.hadoop.hdfs.server.namenode.JournalSet.getInputStream(JournalSet.java:229)
> at org.apache.hadoop.hdfs.server.namenode.FSEditLog.selectInputStreams(FSEditLog.java:1081)
> at org.apache.hadoop.hdfs.server.namenode.ha.EditLogTailer.doTailEdits(EditLogTailer.java:115)
> at org.apache.hadoop.hdfs.server.namenode.ha.EditLogTailer.access$0(EditLogTailer.java:100)
> at org.apache.hadoop.hdfs.server.namenode.ha.EditLogTailer$EditLogTailerThread.run(EditLogTailer.java:154)
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
|