[
https://issues.apache.org/jira/browse/HDFS-13882?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16612971#comment-16612971
]
Xiao Chen commented on HDFS-13882:
----------------------------------
Thanks for the new patch Kitti. 60 secs as the default SGTM. No one really
wants this to grow unlimitedly.
Some comments:
- Should add a test, probably in \{{TestDFSClientRetries}}. I didn't
scrutinize to see if there's a real test on the backoff, but at the minimum we
should add a test similar to
\{{testDFSClientConfigurationLocateFollowingBlockInitialDelay}}.
- {{DFSOutputStream}} had 2 places ({{addBlock}} and {{completeFile}}) that the
retry happens. The new config should cover both, for consistency.
- {{hdfs-default.xml}} default missed a 0.
> Set a maximum for the delay before retrying locateFollowingBlock
> ----------------------------------------------------------------
>
> Key: HDFS-13882
> URL: https://issues.apache.org/jira/browse/HDFS-13882
> Project: Hadoop HDFS
> Issue Type: Improvement
> Affects Versions: 3.1.0
> Reporter: Kitti Nanasi
> Assignee: Kitti Nanasi
> Priority: Major
> Attachments: HDFS-13882.001.patch, HDFS-13882.002.patch
>
>
> More and more we are seeing cases where customers are running into the java
> io exception "Unable to close file because the last block does not have
> enough number of replicas" on client file closure. The common workaround is
> to increase dfs.client.block.write.locateFollowingBlock.retries from 5 to 10.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
