Possibly, you've hitted this: https://issues.apache.org/jira/browse/CASSANDRA-2824 Should be fixed in next minor release.
In the meantime, you "fix" should be alright. -- Sylvain On Wed, Jul 20, 2011 at 3:47 PM, Boris Yen <yulin...@gmail.com> wrote: > Hi Sam, > Thanks for the explanation. > The NodeIds do appear in the Local row of NodeIdInfo, and after manually > deleting two (I got three before I deleted them) of them from CurrentLocal > row, the cassandra can be restarted now. I was just thinking what could be > the possible cause for this? and wondering if anyone has any idea about > this? > Boris > > On Wed, Jul 20, 2011 at 8:00 PM, Sam Overton <sover...@acunu.com> wrote: >> >> The NodeId is used in counter replication. Counters are stored on each >> replica as a set of "shards," where each shard corresponds to the local >> count of one of the replicas for that counter, as identified by the NodeId. >> A NodeId is generated the first time cassandra starts, and might be >> renewed during cleanup, or if you explicitly tell cassandra to generate a >> new one by starting it with -Dcassandra.renew_counter_id >> Do either of the NodeIds in the CurrentLocal row also appear in the Local >> row of the NodeIdInfo CF? the Local row is a history of previous NodeIds >> belonging to this node. If one appears in the history and the other does >> not, it is probably safe to assume that the old NodeId was not deleted from >> the CurrentLocal row correctly, and so you could try removing it from there >> manually. >> >> Sam >> -- >> Sam Overton >> Acunu | http://www.acunu.com | @acunu >> On 20 July 2011 12:25, Boris Yen <yulin...@gmail.com> wrote: >>> >>> Hi, >>> I think we might have screwed our data up. I saw multiple columns inside >>> record: System.NodeIdInfo.CurrentLocal. It makes our cassandra dead forever. >>> I was wondering if anyone could tell me what the NodeId is for? so that I >>> might be able to duplicate this. >>> Thanks in advance >>> Boris > >
