Hello,
We are in the process of preparing our system to run on several clusters. In order to do this we are currently performing some tests, some of which include JBoss cache.
One of these tests is the following:test is the following
1. Having 2 clustered JBoss instances
2. On each instance we deployed the same web application that all it does is increments a cache value for 1000 times
3. Trigger both process together
Expected outcome : The cache value should be 2000
Actual outcome : The value we are getting is somewhere around 1500 - 1600, meaning that we are loosing several updates possibly due to skews
Related configurations in replSync-service.xml :
IsolationLevel - READ_COMMITTED
CacheMode - REPL_SYNC
NodeLockingScheme - MVCC
WriteSkewCheck - true
Software being used :
JBoss - jboss-5.0.0.GA
Jboss Cache - jbosscache-core-3.0.0.GA-all
What might be the problem. Can anybody help us through this ?
Write skew checks only happen with repeatable read.
See http://java.dzone.com/articles/a-look-inside-jboss-cache?mz=3006-jboss for a detailed discussion on how write skews work.