[
https://issues.apache.org/jira/browse/HDFS-15893?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17313852#comment-17313852
]
Ravuri Sushma sree commented on HDFS-15893:
-------------------------------------------
Thank you very much for reviewing this JIRA [~jianghuazhu]
Please correct me if I'm wrong but IIUC you are suggesting if the configured
value of dfs.ha.tail-edits.period is 0 or less than 0, we can set it back to
default Value in order to avoid flooding of logs.
I have added a check of Observer and pushed the logs to debug level because for
observer enabled cluster as it is highly recommended to configure
dfs.ha.tail-edits.period to a very low value (0 ms) so as to decrease the
waiting time of client requests in the RPC queue.
Due to this low value our logs are flooded with any Observer cluster, going by
checking just the value of 0 or below may save us from the flooded logs only if
the configured values are same and may not be very helpful with a value of 0.1
or so. ( which can also cause the same problem).
> Logs are flooded when dfs.ha.tail-edits.in-progress set to true or
> dfs.ha.tail-edits.period to 0ms
> --------------------------------------------------------------------------------------------------
>
> Key: HDFS-15893
> URL: https://issues.apache.org/jira/browse/HDFS-15893
> Project: Hadoop HDFS
> Issue Type: Bug
> Reporter: Ravuri Sushma sree
> Assignee: Ravuri Sushma sree
> Priority: Major
> Attachments: HDFS-15893.001.patch
>
>
> When we set dfs.ha.tail-edits.in-progress to true, dfs.ha.tail-edits.period
> to 0ms almost all the logs on standby and observer NN are loaded. Such logs
> will flood useful logs.
> We can adjust the log level of few logs to debug while observer node is in
> operation.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
