I did a search on the Lucene list archives, found a lot of posts about
the use of Lucene with NFS and how there are locking issues, but don't
see anybody coming to a real solution to this.
We are trying to fix this. Many people seem to hit it.
The current plan is to first decouple the Locking implementation from
the Directory implementation:
http://issues.apache.org/jira/browse/LUCENE-305
and then create a Locking implementation that uses native OS locks
instead of the simple file-existence locks (used now) that do not work
with NFS.
But the development is still in process and so it won't be until a
future Lucene release that this is fixed correctly.
A very good workaround is to use or be inspired by the approach that the
Solr project:
http://incubator.apache.org/solr/features.html
http://incubator.apache.org/solr/tutorial.html
uses, described here:
http://wiki.apache.org/solr/CollectionDistribution
The gist is, your single Java process that has an IndexWriter
periodically snapshots the index at a known-safe time, and then the
multiple Searchers switch to the latest snapshot once the snapshot is
complete.
Solr actually distributes copies of the index to each Searcher's local
storage, but you could probably modify this approach so that, instead,
the Searchers all share a single copy off your SAN.
Mike
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
