hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "mac champion (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-12733) High volume of "Couldn't find oldest seqNum" warnings if bypassing WAL
Date Sun, 27 Mar 2016 17:53:25 GMT

    [ https://issues.apache.org/jira/browse/HBASE-12733?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15213560#comment-15213560
] 

mac champion commented on HBASE-12733:
--------------------------------------

I couldn't find this log statement or any relevant TODOs in FSHLog. Maybe this has been refactored
away already?
Can this be closed, or am I mistaken? 

> High volume of "Couldn't find oldest seqNum" warnings if bypassing WAL
> ----------------------------------------------------------------------
>
>                 Key: HBASE-12733
>                 URL: https://issues.apache.org/jira/browse/HBASE-12733
>             Project: HBase
>          Issue Type: Bug
>            Reporter: Andrew Purtell
>            Priority: Minor
>              Labels: beginner
>
> When ingesting with the WAL bypassed, the RegionServer will log the following at every
flush:
> {noformat}
> 2014-12-19 16:58:24,630 WARN  [MemStoreFlusher.0] wal.FSHLog: Couldn't find oldest seqNum
for the region we are about to flush: [86fc8c1374ebc29cc4a39a78966d48e0]
> {noformat}
> The WAL won't find a seqnum for the region until the first edit which does not bypass
the WAL. That might not be for hundreds or thousands of flushes. This shouldn't be at WARN
level since ingest with WAL bypass is legal if a bit unusual. It's concerning that the FSHLog
code that emits that warning has a TODO above suggesting the warning should instead be an
assertion.



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

Mime
View raw message