hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "stack (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-12978) Region goes permanently offline (WAS: hbase:meta has a row missing hregioninfo and it causes my long-running job to fail)
Date Tue, 10 Feb 2015 18:44:12 GMT

    [ https://issues.apache.org/jira/browse/HBASE-12978?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14314630#comment-14314630
] 

stack commented on HBASE-12978:
-------------------------------

It looks like I broke this:

889333a HBASE-12313 Redo the hfile index length optimization so cell-based rather than serialized
KV key

It used to be this with a context specific comparator:

{code}
byte[] indexKey = comparator.calcIndexKey(lastKeyOfPreviousBlock, firstKeyInBlock);		
dataBlockIndexWriter.addEntry(indexKey, lastDataBlockOffset, onDiskSize);	
{code}

replacing it with....

{code}
165	    // Rather than CellComparator, we should be making use of an Interface here with the
166	    // implementation class serialized out to the HFile metadata. TODO.
167	    Cell indexEntry = CellComparator.getMidpoint(lastCellOfPreviousBlock, firstCellInBlock);
168	    dataBlockIndexWriter.addEntry(CellUtil.getCellKeySerializedAsKeyValueKey(indexEntry),
169	      lastDataBlockOffset, onDiskSize);
{code}

This is not in 0.98.

> Region goes permanently offline (WAS: hbase:meta has a row missing hregioninfo and it
causes my long-running job to fail)
> -------------------------------------------------------------------------------------------------------------------------
>
>                 Key: HBASE-12978
>                 URL: https://issues.apache.org/jira/browse/HBASE-12978
>             Project: HBase
>          Issue Type: Bug
>            Reporter: stack
>             Fix For: 1.0.1
>
>         Attachments: e7cadabc6e5e46c7bf6b3d445f0c53cf
>
>
> Testing 1.0.0 trying long-running tests.
> A row in hbase:meta was missing its HRI entry. It caused the job to fail. Around the
time of the first task failure, there are balances of the hbase:meta region and it was on
a server that crashed. I tried to look at what happened around time of our writing hbase:meta
and I ran into another issue; 20 logs of 256MBs filled with WrongRegionException written over
a minute or two. The actual update of hbase:meta was not in the logs, it'd been rotated off.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message