5 Replies Latest reply on Oct 9, 2013 8:26 AM by iocanel

    zookeeper lost sessions creating lots of logs

    jasonnh

      Hi,

       

      I am using fabric containers on FUSE Enterprise 7.1.0.  I have seen instances when the log files for each child container fill up quickly with thousands of messages like:

       

      ClientCnxn                       | .zookeeper.ClientCnxn$SendThread 1053 | 58 - org.fusesource.fabric.fabric-linkedin-zookeeper - 7.1.0.fuse-047 |Unable to read additional data from server sessionid *****, likely server has closed socket, closing socket connection and attempting reconnect...

       

      The fabric and system in general seems usable but there seems to be 10 (what I assume are zookeeper) sessions that have failed for whatever reason but do not get closed and continually try to reconnect.

       

      Could this be the problem?:

      https://issues.apache.org/jira/browse/ZOOKEEPER-1159

       

      The zookeeper server is running out of connections, e.g.:

      NIOServerCnxnFactory             | 51 - org.fusesource.fabric.fabric-linkedin-zookeeper - 7.1.0.fuse-047 | Too many connections from /A.B.C.D - max is 60

       

      Has anybody else experienced this?

       

      I have about 8 containers and could just up the maximum number of connections but do not know what a reasonably maximum is (60 seems like a lot for 8 containers).

       

      Any help much appreciated.

       

      Thanks.