[
https://issues.apache.org/jira/browse/HDFS-12744?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Weiwei Yang updated HDFS-12744:
-------------------------------
Attachment: HDFS-12744.001.patch
Two line changes in this patch,
1) Added a warning message when SCR is disabled.
2) Log the exception with stack trace to DN log when unknown error happens
while requesting SCR file descriptor, currently only the exception message is
passed to the client without any more logging, causing it's difficult to
analysis what was wrong.
> More logs when short-circuit read is failed and disabled
> --------------------------------------------------------
>
> Key: HDFS-12744
> URL: https://issues.apache.org/jira/browse/HDFS-12744
> Project: Hadoop HDFS
> Issue Type: Improvement
> Components: datanode
> Reporter: Weiwei Yang
> Assignee: Weiwei Yang
> Attachments: HDFS-12744.001.patch
>
>
> Short-circuit read (SCR) failed with following error
> {noformat}
> 2017-10-21 16:42:28,024 WARN
> [B.defaultRpcServer.handler=7,queue=7,port=16020]
> impl.BlockReaderFactory: BlockReaderFactory(xxx): unknown response code ERROR
> while attempting to set up short-circuit access. Block xxx is not valid
> {noformat}
> then short-circuit read is disabled for *10 minutes* without any warning
> message given in the log. This causes us spent some more time to figure out
> why we had a long time window that SCR was not working. Propose to add a
> warning log (other places already did) to indicate SCR is disabled and some
> more logging in DN to display what happened.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: hdfs-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: hdfs-issues-h...@hadoop.apache.org
