[
https://issues.apache.org/jira/browse/HBASE-11240?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
stack updated HBASE-11240:
--------------------------
Fix Version/s: 0.99.0
Release Note: Adds emission of a log message that lists datanodes in HDFS
pipeline when sync is slow. Set when to log with
"hbase.regionserver.hlog.slowsync.ms". Defaults logging if sync takes > 100ms.
Committed to branch-1, both original patch and the addendum. Enis asked that
we do something about too much log when slow. I see this happening in recent
runs. Should be ameliorated by fix to timing calc but my guess is that we will
still have 'too much'. Can address in a follow on patch.
Thanks for the nice addition [~liushaohui]
> Print hdfs pipeline when hlog's sync is slow
> --------------------------------------------
>
> Key: HBASE-11240
> URL: https://issues.apache.org/jira/browse/HBASE-11240
> Project: HBase
> Issue Type: Improvement
> Components: Operability, wal
> Reporter: Liu Shaohui
> Assignee: Liu Shaohui
> Fix For: 0.99.0, 2.0.0
>
> Attachments: HBASE-11240-addendum.diff, HBASE-11240-trunk-v1.diff,
> HBASE-11240-trunk-v2.diff
>
>
> Sometimes the slow sync of hlog writer is because there is an abnormal
> datanode in the pipeline. So it will be helpful to print the pipeline of slow
> sync to diagnose those problems.
> The ultimate solution is to join the trace of HBase and HDFS.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
