recoverBlock timeout in DFSClient should be longer
--------------------------------------------------
Key: HDFS-2378
URL: https://issues.apache.org/jira/browse/HDFS-2378
Project: Hadoop HDFS
Issue Type: Bug
Components: hdfs client
Affects Versions: 0.20.206.0, 0.23.0
Reporter: Todd Lipcon
Assignee: Todd Lipcon
Priority: Critical
Fix For: 0.20.206.0, 0.23.0
In a failure scenario when one of the datanodes in a pipeline has "frozen" (eg
hard swapping or disk controller issues) we sometimes see timeouts in the call
to recoverBlock(). This is because recoverBlock's implementation sends several
RPCs internally (to the NN and to other nodes in the pipeline) with the same
timeout. Since the timeouts are equal, the "outer" call times out first. The
retry then fails since recovery is already in progress, or already finished.
The best fix would be to make recoverBlock idempotent so the retry doesn't
fail, but in the absence of that we can likely fix this issue by increasing the
timeout to be equal to the sum of the timeouts of the underlying recovery calls.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
