Sporadic system load peaks on mail servers using Veritas Cluster File System version 5.0MP4HF2

Issue

• Cluster  of two nodes running RHEL5U4 and Veritas Cluster File  System  (5.0MP4HF2), used as a pop3/imap/smtp mail server with  simultaneuous  read/write access from both nodes on a shared veritas  cluster  filesystem.

• The load average on both systems when operating normally is below 5.

• At   random times, the System Load average rises unexpectidly on one or  both  nodes, and the cause of the load spike cannot be determined.

  • Sometimes   it rises on one node to something in the order of 100 or 200, while it   remains low on the other node. After a few minutes it dissapears and  no  action is required.

  • Sometimes  it rises on both  nodes to something in the order of 1000. It does not  go away after a few  minites and the only way to fix it is to reboot  both nodes  simultaneously

• A case has been opened with Symantec.