Locking issues in cluster, Jackrabbit 2.2.1

[Alvar Lumberg <al...@codeborne.com>]

Hello!


we are using Jackrabbit 2.2.1 for page hierarchy and content storage in a
CMS/portal project we are developing. We have ran into locking issues that
have seemingly appeared/increased after upgrade to 2.2.x from 2.1.3.

We have n instances of our application. Jackrabbit is embedded, configured
to run in cluster, and using DB bundle based shared storage.

The symptoms are that after heavier periods of usage all threads eventually
wait in DefaultISMLock.acquireReadLock(). They are waiting behind a writer
thread that is stuck in AbstractJournal.lockAndSync().

I have seen some related bugfixes introduced in 2.2.x releases. Might our
problem be related with these changes?

A related question that might deserve its own thread. What is the true
function of local repository directory? Is it OK to remove this directory at
will or start new nodes from scratch after old revisions have been purged
from the journal? We have occasionally removed it (not during runtime) and
not experienced anything weird - same for creating new nodes.


With regards
Alvar Lumberg