Looking more in detail I've found that before this error
2007-07-31 01:52:17,917 INFO
virtualdatabase.VirtualDatabaseWorkerThread.elhood
Statement.executeQuery() detected a stored procedure ({call
media_view_insert(267, 1549, 319, 1, 1263553794, null)/* FORCE
REPLICATION */}) remapping the call to CallableStatement.executeQuery()
2007-07-31 01:52:17,920 FATAL backend.DatabaseBackend.node2 Unexpected
error while executing task: null task
java.lang.NullPointerException
at
org.continuent.sequoia.controller.requests.StoredProcedure.parse(StoredProcedure.java:488)
at
org.continuent.sequoia.controller.loadbalancer.BackendTaskQueues.fetchNextQueryFromBackendTotalOrderQueue(BackendTaskQueues.java:837)
at
org.continuent.sequoia.controller.loadbalancer.BackendTaskQueues.getNextEntryToExecute(BackendTaskQueues.java:1693)
at
org.continuent.sequoia.controller.loadbalancer.BackendWorkerThread.run(BackendWorkerThread.java:185)
the {call .. request was written in node1 log but not in node2, seems
that the controller was not able to deliver the request to node2.
Yesterday this same stored procedure (using the same parameters) was
executed mostly successfuly, but failed a couple of times with "null
task" error under high load, leaving open transactions in the
controller. Therefore I was not able to kill those transactions with
"abort transaction".
This error is turning the cluster inconsistent so I'm working right now
with one node on my production environment.
How can I debug and fix this problem ?
Thanks,
Saulo G.
Hi all,
Using Sequoia 2.10.8 with one controller and two PostgreSQL nodes I
have the following issue:
Node2 was having problems processing (only some) requests, and the
controller logged:
2007-07-31 01:52:17,917 INFO
virtualdatabase.VirtualDatabaseWorkerThread.elhood
Statement.executeQuery() detected a stored procedure ({call
media_view_insert(267, 1549, 319, 1, 1263553794, null)/* FORCE
REPLICATION */}) remapping the call to CallableStatement.executeQuery()
2007-07-31 01:52:17,920 FATAL backend.DatabaseBackend.node2 Unexpected
error while executing task: null task
java.lang.NullPointerException
at
org.continuent.sequoia.controller.requests.StoredProcedure.parse(StoredProcedure.java:488)
at
org.continuent.sequoia.controller.loadbalancer.BackendTaskQueues.fetchNextQueryFromBackendTotalOrderQueue(BackendTaskQueues.java:837)
at
org.continuent.sequoia.controller.loadbalancer.BackendTaskQueues.getNextEntryToExecute(BackendTaskQueues.java:1693)
at
org.continuent.sequoia.controller.loadbalancer.BackendWorkerThread.run(BackendWorkerThread.java:185)
Every failed transactions (with "null task") after this point remained
open from the controller point of view.
vdb(admin) > dump scheduler queues
Active transactions: 0
Pending read requests: 1
Read request id list:
322181
Pending write requests: 5
Write request id list:
32385 266795 70499 23932 22434
all those pending transactions were executed sucessfully on node1 but
not in node2 (leaving the cluster inconsistent), and the controller
logged:
2007-07-31 22:12:19,369 WARN sequoia.controller.scheduler Waiting for
4 pending writes
2007-07-31 22:16:08,527 WARN sequoia.controller.scheduler Waiting for
4 pending writes
2007-07-31 23:40:43,133 WARN sequoia.controller.scheduler Waiting for
4 pending writes
I understand that the controller should abort and rollback all these
transactions. Is this right?
The relevant parts of muy virtualdatabase conf are:
<RequestScheduler>
<RAIDb-1Scheduler level="passThrough"/>
</RequestScheduler>
<RequestCache>
<MetadataCache />
<ParsingCache />
<ResultCache granularity="column" pendingTimeout="2">
<DefaultResultCacheRule >
<NoCaching />
</DefaultResultCacheRule>
</ResultCache>
</RequestCache>
<LoadBalancer>
<RAIDb-1>
<WaitForCompletion policy="all"/>
<RAIDb-1-LeastPendingRequestsFirst/>
</RAIDb-1>
</LoadBalancer>
Any help or direction on this matter would by highly appreciated
Regards,
Saulo G.
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
