-
1. Re: Issues when adding a new node into a running cluster
changgeng Jan 4, 2011 8:27 PM (in response to changgeng)Any idea on this issue? Are we going to address it in the next release?
-
2. Re: Issues when adding a new node into a running cluster
an1310 Jan 12, 2011 7:50 AM (in response to changgeng) -
3. Re: Issues when adding a new node into a running cluster
manik Jan 12, 2011 7:52 AM (in response to an1310)I made a comment on the JIRA; I'll repeat here.
I think this may be the culprit - https://github.com/infinispan/infinispan/blob/master/core/src/main/java/org/infinispan/interceptors/DistributionInterceptor.java#L247
Happens when a prepare() occurs on, say, nodes {A, B, C} and the commit is sent to {A, B, D} since D joins between the prepare and commit, and D takes ownership of the key.
Do you see this exception as a signature of this failure, prior to seeing timeout exceptions?
java.lang.IllegalStateException: Can not commit since DldGlobalTransaction{coinToss=NNNNN, isMarkedForRollback=false, lockIntention=null, affectedKeys=[], locksAtOrigin=[K]} GlobalTransaction:<address>:port:local was prepared on [C1, C2, C3] nodes while it is being committed to [C1, C2, C4]
Also, I haven't been able to consistently reproduce the issue - so help would be appreciated if you can consistently reproduce it with your test, using TRACE level logging for Infinispan.