with some help from this list, I’ve finally managed to get a clustered Jackrabbit test setup running. So far everything seems good, but I’ve run into an issue I’ve been expecting all along.
Say two users U and V are accessing cluster nodes A and B. If U writes resource X on A and immediately afterwards V writes the same resource X on B, it can happen that B’s changes are gone immediately after they’re written and replaced by A’s changes. I suppose this happens when both writes happen within the same synchronisation time frame (I left syncDelay at the default 5000ms).
As far as I understand, the section “Concurrent Write Behavior” at the end of the clustering wiki page seems to imply that locking JCR nodes before writing to them forces a cluster sync. In this case, the behaviour I’m seeing would imply that I either still have a configuration issue, or that Jackrabbit’s WebDAV layer simply doesn’t use locking and therefore doesn’t prevent these concurrency issues.
I’d appreciate feedback and comments on this, if only to confirm that this is indeed working as designed and I can stop worrying about my configuration.
DI Marian Schedenig
Qualysoft GmbH | Saturn Tower, Leonard-Bernstein-Straße 10, A-1220 Wien | Fimenbuchnummer 186076t, Handelsgericht Wien
Austria - Germany - Hungary - Romania - Serbia – Slovakia - Ukraine
P Please consider the environment before printing this email