Late reply, but I just got the same error restarting after upgrading from 0.7.2 to 0.7.5.
I did a drain using nodetool on each node before I killed them and did the upgrade. Should all commitlogs have been cleaned up after a drain? I would think so, but they were not. Maybe there is a bug around not cleaning up certain stale commitlogs, which, when replayed causes the error. On Fri, Apr 29, 2011 at 2:02 PM, aaron morton <aa...@thelastpickle.com>wrote: > I've not heard of it happening in a production system. > > The commit log and the schema have diverged, that does not normally > happen. > > Aaron > > On 29 Apr 2011, at 20:54, Subscriber wrote: > > But isn't their a better way than to delete the commitlogs? > What do I do when this happens in a productive system? > > Ciao > Udo > > Am 29.04.2011 um 02:07 schrieb aaron morton: > > Thought you may have re-created the schema. > > Kill the process like that should be ok, let us know if you get the error > again. > > Aaron > > On 29 Apr 2011, at 02:56, Subscriber wrote: > > Hi Aaron, > > what exactly do you mean? > I restarted the cluster by calling > > > bin/cassandra -p pid.file > > on all three nodes. The first node is the (only) seed. > > Udo > > > Am 27.04.2011 um 23:28 schrieb Aaron Morton: > > What approach did you take to restarting the cluster? > > It looks like the keyspace name was changed and the log replay tried to > write to the old one. > > Aaron > > On 28/04/2011, at 12:03 AM, Subscriber <subscri...@zfabrik.de> wrote: > > Hi again, > > some more remarks. > I renamed the commitlog directory on the third node so that cassandra > cannot see it on startup. > Now the node starts fine. The problem seems to have something to do with > the commitlogs... > > Best Regards > Udo > > > Am 27.04.2011 um 13:22 schrieb Subscriber: > > Hi, > > I'm using Cassandra 0.7.4 on a three node cluster. The cluster was setup > yesterday as a fresh installation (no upgrade). > The cluster is installed beside a hadoop cluster (I want to discover how > cassandra works together with hadoop's map/reduce feature). > > After loading some test data into the cassandra cluster (some text-files > for a first wordcount m/r example ;-) I killed all nodes and restarted the > cluster. > Two nodes are starting fine but the third node doesn't start and throws a > NullPointerException: > > INFO 12:44:42,557 Finished reading > /home/.../cassandra/work/commitlog/CommitLog-1303828029742.log > ERROR 12:44:42,558 Exception encountered during startup. > java.lang.NullPointerException > at org.apache.cassandra.db.Table.createReplicationStrategy(Table.java:275) > at org.apache.cassandra.db.Table.<init>(Table.java:221) > at org.apache.cassandra.db.Table.open(Table.java:110) > at org.apache.cassandra.db.commitlog.CommitLog.recover(CommitLog.java:273) > at org.apache.cassandra.db.commitlog.CommitLog.recover(CommitLog.java:156) > at > org.apache.cassandra.service.AbstractCassandraDaemon.setup(AbstractCassandraDaemon.java:173) > at > org.apache.cassandra.service.AbstractCassandraDaemon.activate(AbstractCassandraDaemon.java:314) > at > org.apache.cassandra.thrift.CassandraDaemon.main(CassandraDaemon.java:79) > Exception encountered during startup. > > This looks pretty much the same as issue > <https://issues.apache.org/jira/browse/CASSANDRA-1889> > https://issues.apache.org/jira/browse/CASSANDRA-1889 which is in > resolution "cannot reproduce". > > What can I do to make the node starting again? > > Thanks in advance and best regards > Udo > > > > > > >
