hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Enis Soztutar (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-8205) HBCK support for table locks
Date Thu, 28 Mar 2013 18:39:17 GMT

    [ https://issues.apache.org/jira/browse/HBASE-8205?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13616515#comment-13616515
] 

Enis Soztutar commented on HBASE-8205:
--------------------------------------

Thanks Jon for taking a look,
bq. This seems out of place for the generic lock interface. Can you explain why this changed?
To reflect what the function actually does. It only deletes write locks. 
bq. javadoc: this is both the read and write aspects of the table lock right?
Yes, if it does not say read or write explicitly, it means all locks. 
bq. javadoc here says the value must not be null[1]. Why is this lock.writeLock(null) ok?
We are passing null, because we won't attempt to acquire() the lock. So it is fine. Metadata
is serialized inside the znode. 
bq. comment nit: double negatives are complicated
Should have read: "should not have expired, no problems" (notice the added ","). Added it
in the next version
bq. Need to explain the false cases in javadoc.
done
bq. help me here – what does the magic value mean?
Zk SEQUENTIAL nodes contain 10-digit sequence number appended, which is explained in ZKInterProcessLockBase.ZNodeComparator.getChildSequenceId().
This sends the largest sequence value. I'll add some more comments in the line. 
bq. Put this in the interface and then inheritdoc + add the zk data here?
Those interfaces are not in a hierarchy. TableLockManager.reapExpiredLocks() uses ZKInterProcessWriteLock.reapExpiredLocks()
internally. 

I've put up the patch in RB, should you have any further comments.
                
> HBCK support for table locks
> ----------------------------
>
>                 Key: HBASE-8205
>                 URL: https://issues.apache.org/jira/browse/HBASE-8205
>             Project: HBase
>          Issue Type: Improvement
>          Components: hbck, master, regionserver
>            Reporter: Enis Soztutar
>            Assignee: Enis Soztutar
>             Fix For: 0.95.0, 0.98.0
>
>         Attachments: hbase-8205_v1.patch, hbase-8205_v2.patch
>
>
> Table locks have been introduced in HBASE-7305, HBASE-7546, and others (see the design
doc at HBASE-7305). 
> This issue adds support in HBCK to report and fix possible conditions about table locks.
Namely, if due to some bug, the table lock remains not-released, then HBCK should be able
to report it, and remove the lock, so that normal table operations will continue. 
> Also see the comments in HBASE-7977. 

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message