Problems with upgraded version of ehCache (1.3).

(In [3082]) Upgrade ehcache to 1.3.0.

Also now the peer-listeners are by default configured to listen on TCP port 8111 for plosone webapp. Note that the cache-invalidate filter-resolver on mulgara-server does not listen for updates from peer. It only notifies peers on updates. Therefore for firewalls do the following:

allow multicast ---> plosone, mulgara allow rcv on tcp 8111 --> plosone allow send to tcp 8111 --> plosone, mulgara

This addresses #426

After upgrade to 1.3 we're seeing the following problems:

• excessive logging on mulgara - can be temporarily fixed with the following entry in commons/classes/log4j.xml:

    <logger name="net.sf.ehcache">
        <level value="warn" />
    </logger>
  

• excessive cpu usage on plosone server (cpu constantly maxed out)

This source of problem has been identified and a bug report (including patch) has been submitted: https://sourceforge.net/tracker/index.php?func=detail&aid=1755263&group_id=93232&atid=603559 .

The excessive logging and CPU hogging are from the same source; the excessive logging can be cured by setting the log-level for net.sf.ehcache.distribution to info instead of debug.

In the mean time, the only solution to the CPU hogging is to either disable the cacheManagerPeerProviderFactory in ehcache.xml, or to ensure that multicast packets are allowed through the firewall (dropping them instead of rejecting them is not sufficient).

(In [3279]) Use a patched ehcache.jar deployed on gandalf. The patch addresses excessive cpu load on exception during multi-cast-heart-beat sends. (exceptions by-pass a sleep(). The patch makes sure sleep() is called always.

Ronald has submitted a patch (see https://sourceforge.net/tracker/index.php?func=detail&aid=1755263&group_id=93232&atid=603559)

Closes #426.

Milestone 0.8 deleted