Right, as I mentioned in my first post, I ran nodeprobe flush first, then rolled forward. However, I checked and the commitlogs were non-zero so maybe it didn't take. A roll back to 0.4.2 and a restart caused the commit logs to get flushed, then I reran nodeprobe flush again, then checked commit log sizes and restarted. Seems like it's doing something now, and not running out of memory, so hopefully it worked.
Next time I run, I'll call nodeprobe flush and check commit log sizes prior to shutting instances down, then maybe it won't be an issue. -Anthony On Tue, Jan 19, 2010 at 05:40:27PM -0500, Ray Slakinski wrote: > The commit log format is different from 0.4.2 to 0.5.0, you might want to > roll back to 0.4.2 and do a nodeprobe flush, then upgrade to 0.5.0. If you > dont need the commit log data you can always remove it before starting 0.5.0, > but only do so if your ok with possibly loosing some data. > > Ray Slakinski > > On 2010-01-19, at 5:33 PM, Anthony Molinaro wrote: > > > Hi, > > > > I just tried upgrading a set of machines from 0.4.2 to 0.5.0 final, > > and the following happens on startup > > > > java.lang.OutOfMemoryError: Java heap space > > at org.apache.cassandra.db.CommitLog.recover(CommitLog.java:318) > > at > > org.apache.cassandra.db.RecoveryManager.doRecovery(RecoveryManager.java:65) > > at > > org.apache.cassandra.service.CassandraDaemon.setup(CassandraDaemon.java:90) > > at > > org.apache.cassandra.service.CassandraDaemon.main(CassandraDaemon.java:166) > > > > Prior to attempting the upgrade, I removed all traffic from these instances, > > ran cassandra-nodeprobe flush on each, then shut all down. Upon starting > > the first instance I got the above. Any ideas? > > > > -Anthony > > > > -- > > ------------------------------------------------------------------------ > > Anthony Molinaro <antho...@alumni.caltech.edu> > -- ------------------------------------------------------------------------ Anthony Molinaro <antho...@alumni.caltech.edu>
