Infinispan 5.1.4.Final + JGroups 3.0.9.Final
bramalingam81 May 24, 2012 7:46 AMHi,
We are trying to setup Infinispan Cluster on 3 Nodes with Distributed Clustering mode. We simulated a load of 100,000 request and after around 20,000 request, all the nodes got hanged and it was unable to take up any more request. We simulated this load using jmeter scripts. Data that is being pushed to each of the nodes is of 12kb. Our expectation was to have the replication happen as immediately as an entry gets added/updated in the cache.
However, it was NOT happening. Other scenario that we tested was to have only 2 nodes currently serving the load mentioned above, and when we brought the 3rd node, NO nodes were NOT able to take up any request. All the nodes got blocked.
While I was searching for the way the Distributed Cache works, I found the following two links to understand more on the distributed mode.
https://community.jboss.org/wiki/DIST-DistributedCacheMode
https://community.jboss.org/thread/196059?start=0&tstart=0
Please help me in validating my configuration files attached.
1. We do NOT want the cache to be persistent.
2. All data that goes into the cache will be String. Instead of relying on Java Serialization, we have convered all the objects to a json String and pushed the same to cache.
I have attached 2 files : infinispan & jgroups configuration file.
In JGroups - We made it to do a streaming state transfer. So enabled the same by adding this protocol.
<pbcast.STATE max_pool="100" pool_thread_keep_alive="1800000" buffer_size="81920"/>
In infinispan - We configured for asynchronous state transfer
<namedCache name="application-cache">
<eviction strategy="LIRS" maxEntries="1800000" />
<expiration maxIdle="36000" wakeUpInterval="1000" />
<clustering mode="distribution">
<stateTransfer fetchInMemoryState="false" timeout="20000" />
<async asyncMarshalling="false" useReplQueue="true"
replQueueInterval="10" replQueueMaxElements="1000" />
<l1 enabled="true" lifespan="36000" onRehash="true" />
<hash numVirtualNodes="9" numOwners="3" />
</clustering>
<loaders passivation="false" shared="false" preload="true">
<loader class="org.infinispan.loaders.file.FileCacheStore"
fetchPersistentState="true" ignoreModifications="false"
purgeOnStartup="true">
<properties>
<property name="location" value="/home/balaji/" />
</properties>
</loader>
</loaders>
<transaction useEagerLocking="true" eagerLockSingleNode="true">
<recovery enabled="true" recoveryInfoCacheName="noRecovery" />
</transaction>
</namedCache>
Can you please validate the configurations to know on any flaw in the same.
-
jgroups-app.xml 2.0 KB
-
infinispan-cluster.xml 3.2 KB