Still no idea on the OOM - please send the stacktrace if you can. As for doing a replication recovery when it should not be necessary, yonik just committed a fix for that a bit ago.
On Aug 7, 2012, at 9:41 AM, Mark Miller <markrmil...@gmail.com> wrote: > > On Aug 7, 2012, at 5:49 AM, Jam Luo <cooljam2...@gmail.com> wrote: > >> Hi >> I have big index data files more then 200g, there are two solr >> instance in a shard. leader startup and is ok, but the peer alway OOM >> when it startup. > > Can you share the OOM msg and stacktrace please? > >> The peer alway download index files from leader because >> of recoveringAfterStartup property in RecoveryStrategy, total time taken >> for download : 2350 secs. if data of the peer is empty, it is ok, but the >> leader and the peer have a same generation number, why the peer >> do recovering? > > We are looking into this. > >> >> thanks >> cooljam > > - Mark Miller > lucidimagination.com > > > > > > > > > > > - Mark Miller lucidimagination.com