[
https://issues.apache.org/jira/browse/HADOOP-1990?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12532198
]
stack commented on HADOOP-1990:
-------------------------------
Looking at TDFSA, as best as I can tell w/o thread dumps, it looks like the
HADOOP-1966 fail. Exceptions logged are same in about same order. In
HADOOP-1996, the thread dumps have us hung in a flush down in the IPC Client
preventing the cluster's clean exit.
This hang happens sporadically but w/ some regularity on Hudson. Suggest
backing out TDFSA for a while. We can add it back in later after a spate of
successful build runs and after addition of new debugging/ fix-attempts.
> [hbase] Nightlies #258 and #259 failed
> --------------------------------------
>
> Key: HADOOP-1990
> URL: https://issues.apache.org/jira/browse/HADOOP-1990
> Project: Hadoop
> Issue Type: Bug
> Components: contrib/hbase
> Reporter: stack
> Assignee: Jim Kellerman
>
> Looking at the failures in the last two nightlies.
> Looking at #258, the test that failed was same as reported in HADOOP-1964 but
> mode is this time different. The ROOT scanner keeps running but the META has
> disappeared with a log emission. Its probably hung. A misreported missing
> filesystem LOG info (bug) gives wrong impression that the FS has gone. Jim's
> HADOOP-1966 has changed how things work significantly enough that am going to
> pass on further analysis of this failure.
> On #259, Jim has handle on two of the three failed tests. I'm looking at the
> TestDFSAbort...
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
