[
https://issues.apache.org/jira/browse/HBASE-1634?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12730669#action_12730669
]
stack commented on HBASE-1634:
------------------------------
Haijun Cao are you on TRUNK or on 0.19.x hbase? If on TRUNK, it would be
interesting to see if you have HBASE-1638 condition. Otherwise, its something
else and should be self-healing.
Update to latest TRUNK to pick up some fixes. One of the fixes prints out the
name of the row that is provoking the NPE. Try deleting that row from .META.
table from the shell.
Does that fix it?
If not, may be HBASE-1638.
You can figure out the file with bad entry by doing something like the
following:
$ ./bin/hbase org.apache.hadoop.hbase.io.hfile.HFile
/tmp/hbase-stack/hbase/.META./1028785192/info/EACH_FILE_UNDER_INFO |grep
historian
... substituting your filenames in the above and your path to your hbase
install.
This will emit a row if has historian family entry. If you see one, then you
have HBASE-1638. Somethings up.
https://issues.apache.org/jira/browse/HBASE-1638 has a fixup script for the
file with the bad entry in it. See issue for how to run it.
After amending the file in .META., redeploy meta or just restart the cluster.
> UI and shell list throw NPE
> ---------------------------
>
> Key: HBASE-1634
> URL: https://issues.apache.org/jira/browse/HBASE-1634
> Project: Hadoop HBase
> Issue Type: Bug
> Reporter: stack
> Fix For: 0.20.0
>
>
> I have a .META. from bryan duxbury that has a row with only historian data.
> List does whole row, not just info, so if only historian entries, a row is
> returned. We then try and get a regioninfo from this row and NPE.
> As to why this happens, jd suggests that since we moved to async historian
> updates, row may have been deleted already. Confirm.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
