After further investigation by David and myself, it appears that on networked file systems such as NFS, performance depends on the speed of the network as well as the NFS implementation. It has been observed that file locking over NFS is biased so that the current owner of the lock is favored over other processes. Thus, while one process hogs the lock for the log file, other processes starve waiting for the lock to the point of appearing deadlocked.

The impact of prudent mode is highly dependent on network speed and OS implementation details. We provide an very small application called FileLockSimulator[1] which can help to simulate the behavior of prudent mode in your environment.

David, would you concur with the above or want to add anything?

[1] https://gist.github.com/2794241

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators.
For more information on JIRA, see: http://www.atlassian.com/software/jira