[
https://issues.apache.org/jira/browse/HBASE-10048?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13836177#comment-13836177
]
chunhui shen commented on HBASE-10048:
--------------------------------------
bq,couldn't just we make an intelligent "guess" about log size rather than
creating a separate metrics for the same.
I know the hlog size is about 95% of block size in normal case,
but we have seen the hlog file size much bigger than block size in abnormal
case.
I think such a metric could show the unhealthy scenario timely, avoiding the
hlog size become very huge.
Maybe we only need the metric for the max size of current hlog.
[~liushaohui] Any thought about the log size metric?
> Add hlog number metric in regionserver
> --------------------------------------
>
> Key: HBASE-10048
> URL: https://issues.apache.org/jira/browse/HBASE-10048
> Project: HBase
> Issue Type: Improvement
> Reporter: Liu Shaohui
> Assignee: Liu Shaohui
> Priority: Minor
> Attachments: HBASE-10048-0.94-v1.diff, HBASE-10048-0.94-v2.diff,
> HBASE-10048-trunk-v1.diff, HBASE-10048-trunk-v2.diff,
> HBASE-10048-trunk-v3.diff
>
>
> Add hlog number metric in regionserver.
> We can use this metric to alert about memstore flush because of too many
> hlogs.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
