[
https://issues.apache.org/jira/browse/HBASE-13452?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14496711#comment-14496711
]
Enis Soztutar commented on HBASE-13452:
---------------------------------------
Agreed, the warning message should be more descriptive. In cases where the
memstore size is not 0, we log it, since it may indicate a data loss. It can
either be that the size calculation is off, or we skip memstore flush in
close(). If the region server is aborting we skip the memstore flush for
example. Is it the case?
> HRegion warning about memstore size miscalculation is not actionable
> --------------------------------------------------------------------
>
> Key: HBASE-13452
> URL: https://issues.apache.org/jira/browse/HBASE-13452
> Project: HBase
> Issue Type: Bug
> Affects Versions: 1.0.0
> Reporter: Dev Lakhani
> Priority: Blocker
> Fix For: 2.0.0, 1.1.0, 1.0.2
>
>
> During normal operation the HRegion class reports a message related to
> memstore flushing in HRegion.class :
> if (!canFlush) {
> addAndGetGlobalMemstoreSize(-memstoreSize.get());
> } else if (memstoreSize.get() != 0) {
> LOG.error("Memstore size is " + memstoreSize.get());
> }
> The log file is filled with lots of
> Memstore size is: 190192
> Memstore size is: 442232
> Memstore size is: 190192
> ...
> These message are uninformative, clog up the logs and offers no root cause
> nor solution. Maybe the message needs to be more informative, changed to WARN
> or some further information provided.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
