3 messages in org.apache.jackrabbit.usersRe: Rebuild Lucene Index For Clustere...
FromSent OnAttachments
Noah VihinenOct 12, 2007 8:30 am 
Noah VihinenOct 18, 2007 6:57 am 
Dominique PfisterOct 18, 2007 7:04 am 
Actions with this message:
Paste this link in email or IM:
Paste this link in email or IM:
Atom feed for this thread
Paste this URL into your reader:
Subject:Re: Rebuild Lucene Index For Clustered NodeActions...
From:Noah Vihinen (nvih@maven.net)
Date:Oct 18, 2007 6:57:34 am
List:org.apache.jackrabbit.users

I really wish someone could confirm my findings, as there doesn't seem to be documentation on this. If I've missed the documentation, can someone please point me to it?

Our current understanding is this. No matter what, when search indexes are removed on a clustered jackrabbit node, the indexes are rebuilt from the persistent node/property/value states in the DB, ignoring the journal table. Following that, everything from the local journal revision number (in the revision file) up to the latest existing revision is replayed. If the revision file has been removed, a new one is created with revision 0, which results in every entry in the journal table being replayed.

Is all of this correct?

Thanks, Noah

On Oct 12, 2007, at 11:30 AM, Noah Vihinen wrote:

I want to confirm the behavior of lucene search index rebuilding in a clustered environment. Our clustered environment is configured to use the database journal table to keep the nodes in sync. When one node rebuilds its lucene search index, does it rebuild it by replaying every relevant change in the journal table?

In our experience this is true. Assuming this is the case, is there any way to 'reset' the journal table, and make the lucene search indexes rebuild off of the actual persistent node/property/ vale state in the DB, rather than the journal table entries?

Thanks, Noah