Details
-
Bug
-
Status: Open
-
Major
-
Resolution: Unresolved
-
1.6.0
-
None
-
None
Description
I have a long running service which uses HDFS sink to push data to a Kerberos-enabled HDFS cluster.
However, the HDFS sink always fail after 10 hours running saying something like:
16 Nov 2015 10:32:54,400 WARN (org.apache.hadoop.security.UserGroupInformation.hasSufficientTimeElapsed:1168) - Not attempting to re-login since the last re-login was attempted less than 600 seconds before.
And then later followed by some " GSS initiate failed".
I checked the FlumeAuthenticator and found that there's a method "startCredentialRefresher" which will spawn a new thread to do relogin.
I believe this method should be called in the HDFSEventSink.