The documentation could maybe be improved, but the basics of backup snapshots with the in-process (Java-based) replication handler actually seem pretty straightforward to me, now that I understand it:
1. You can make a snapshot whenever you want by hitting http://master_host:port/solr/replication?command=backup 2. You can have automatically triggered snapshots at commit time or optimize time by putting a backupAfter tag in the replication handler section of your solrconfig.xml. (See http://wiki.apache.org/solr/SolrReplication) In neither case do you need to stop Solr or stop modifying your index while the backup is in progress. Does anything in particular seem not straightforward? I guess there's no built-in way to purge old indexes from disk; that's a little inconvenient. If you want to use the command-line tools, I think those should be totally compatible with the new (Java) replication tools. I don't know as much about them, though. 2009/9/29 KaktuChakarabati <jimmoe...@gmail.com>: > > Hey, > I noticed with new in-process replication, it is not as straightforward to > have > (production serving) solr index snapshots for backup (it used to be a > natural byproduct > of the snapshot taking process.) > I understand there are some command-line utilities for this (abc..) > Can someone please explain how to use these to take a snapshot > of a solr index, assuming it is being used in production? what are some > guidelines? should I stop > other processes that might be issuing updates and/or comitts while taking it > or is it atomic (e.g hard link )? > > would be nice to have this in wiki too i think for the benefit of other > users, > having regular backup snapshots seems critical.. > > Thanks, > -Chak > -- > View this message in context: > http://www.nabble.com/Index-backup-with-new-replication--tp25667145p25667145.html > Sent from the Solr - User mailing list archive at Nabble.com. > >
