shuffler penalty box delay time can be negative
-----------------------------------------------
Key: HADOOP-5984
URL: https://issues.apache.org/jira/browse/HADOOP-5984
Project: Hadoop Core
Issue Type: Bug
Reporter: Aaron Kimball
We encountered some syslogs with the following lines:
"attempt_attempt_20090621547_0037_r_000000_0 adding host ftrp0cmt0095 to
penalty box, next contact in 4 seconds"
"attempt_attempt_20090621547_0037_r_000000_0 adding host ftrp0cmt0095 to
penalty box, next contact in 8 seconds"
"attempt_attempt_20090621547_0037_r_000000_0 adding host ftrp0cmt0095 to
penalty box, next contact in 16 seconds"
"ftrp-cmt0095 Will be considered after -3 seconds"
"ftrp-cmt0095 Will be considered after -63 seconds"
"ftrp-cmt0095 Will be considered after -123 seconds"
"ftrp-cmt0095 Will be considered after -183 seconds"
I tried tracing through the relevant code in o.a.h.m.ReduceTask.fetchOutputs()
but it's got a pretty lengthy control flow. Does anyone know why the
'currentTime' field should skew so much?
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
