[
https://issues.apache.org/jira/browse/HBASE-16661?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15518096#comment-15518096
]
Gary Helmling commented on HBASE-16661:
---------------------------------------
No, I think ERROR level is fine in this case. FATAL is usually reserved for
cases where we cannot continue, implying the regionserver should abort.
Instead of -1, I would interpret the exception (or no major compaction at all)
as a timestamp of 0, so age in that case will be the same as currentTime().
It's true that that doesn't let us distinguish between an error and no major
compaction at all, but a -1 doesn't help if we're doing things like checking
for age > some threshold for alerting.
> Add last major compaction age to per-region metrics
> ---------------------------------------------------
>
> Key: HBASE-16661
> URL: https://issues.apache.org/jira/browse/HBASE-16661
> Project: HBase
> Issue Type: Improvement
> Reporter: Gary Helmling
> Assignee: Dustin Pho
> Priority: Minor
> Attachments: HBASE-16661.001.patch
>
>
> After HBASE-12859, we can now track the last major compaction timestamp for
> each region. However, this is only exposed through cluster status reporting
> and the admin API.
> We have similar per-region metrics around storefile age, but none that
> filters on major compaction specifically.
> Let's add a metric for last major compaction age.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
