[jbosscache-dev] LockParentForChildInsertRemove and PessimisticLocking
Brian Stansberry
brian.stansberry at redhat.com
Wed Aug 5 12:03:44 EDT 2009
Just to be more thorough, these tests involve a 4 node cluster. Test
driver spawns 400 client threads which continually call a servlet,
incrementing a counter stored in the session. Load across the cluster
when all nodes are running is about 190 req / second.
The test cycles through {A, B, C, D} first shutting down A, running 1
minute then restarting A, running one minute, shutting down B, running,
restarting, and so on. In this case C had just been shut down; logging
is from B.
Brian Stansberry wrote:
> Node 10.34.32.155 had been shutdown cleanly (i.e. not a hard kill) about
> 9 seconds before the logging I posted.
>
> At the time of the problematic gravitation there is a lot of gravitation
> going on, not just this one thread.
>
> Manik Surtani wrote:
>> Interestingly, the cleanup command is for a dead backup. Does your test
>> involve a gravitation from a backup of a dead node, e.g., {A, B, C, D}
>> and A is a data owner. A dies and a lookup is done on D?
>>
>> Cheers
>> Manik
>>
>> On 5 Aug 2009, at 14:42, Brian Stansberry wrote:
>>
>>> DataGravitationCleanupCommand{fqn=/JSESSION/st_localhost/wUaU4FhUzWcuiDEjkJzDiQ__,
>>>
>>> backup=/_BUDDY_BACKUP_/10.34.32.155_41754:DEAD/1/JSESSION/st_localhost/wUaU4FhUzWcuiDEjkJzDiQ__}
>>>
>> --
>> Manik Surtani
>> manik at jboss.org
>> Lead, Infinispan
>> Lead, JBoss Cache
>> http://www.infinispan.org
>> http://www.jbosscache.org
>>
>>
>>
>>
>
>
--
Brian Stansberry
Lead, AS Clustering
JBoss by Red Hat
More information about the jbosscache-dev
mailing list