Hi we are thinking of implementing a solution like this. All nodes on the cluster has a local index from which searches are performed. Once an update on the index needs to happen the work is sent to a queue where there is a master who is responsible for updating the central index. The master is then responsible for propagating/notifying other nodes in the cluster that their local index is not in sync. The local index worker can perform an update against the master index.
We haven't implemented this yet but its the approach we'll take inorder to deal with clustering. If anyone has any ideas which could improve this or a better approach your thoughts will be appreciated. Cheers Amin On Wed, Jun 10, 2009 at 10:15 AM, Ian Lea <ian....@gmail.com> wrote: > I'd recommend using your favourite queueing service to pass all > updates to a central process, the one and only process that updates > the index. If you don't already have a favourite queueing service, > http://en.wikipedia.org/wiki/Java_Message_Service#Provider_implementations > lists several JMS implementations. > > > -- > Ian. > > On Tue, Jun 9, 2009 at 9:02 PM, Newman, Billy<billy.new...@itt.com> wrote: > > I am trying to figure out the best way to add to a lucene index across a > clustered app server. I cannot grab an IndexWriter for each node in the > cluster, because I would run into lock file problems. I am not sure if I > can share one IndexWriter across the cluster because what happens when two > or more nodes update the same thing. Anyone have any insight? > > > > Thanks. > > --------------------------------------------------------------------- > To unsubscribe, e-mail: java-user-unsubscr...@lucene.apache.org > For additional commands, e-mail: java-user-h...@lucene.apache.org > >
