hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "He Liangliang (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-13200) Improper configuration can leads to endless lease recovery during failover
Date Thu, 12 Mar 2015 01:59:39 GMT

    [ https://issues.apache.org/jira/browse/HBASE-13200?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14357973#comment-14357973
] 

He Liangliang commented on HBASE-13200:
---------------------------------------

Yes, we encountered this issue in prod then make this fix.
This fix is aimed to avoid the worst case, although it means something else might be wrong
if it reaches the 3rd retry.

> Improper configuration can leads to endless lease recovery during failover
> --------------------------------------------------------------------------
>
>                 Key: HBASE-13200
>                 URL: https://issues.apache.org/jira/browse/HBASE-13200
>             Project: HBase
>          Issue Type: Bug
>          Components: MTTR
>            Reporter: He Liangliang
>            Assignee: He Liangliang
>         Attachments: HBASE-13200.patch
>
>
> When a node (DN+RS) has machine/OS level failure, another RS will try to do lease recovery
for the log file. It will retry for every hbase.lease.recovery.dfs.timeout (default to 61s)
from the second time. When the hdfs configuration is not properly configured (e.g. socket
connection timeout) and without patch HDFS-4721, the lease recovery time can exceeded the
timeout specified by hbase.lease.recovery.dfs.timeout. This will lead to  endless retries
and preemptions until the final timeout.



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

Mime
View raw message