-
Bug
-
Resolution: Done
-
Critical
-
8.0.0.Beta-CR1
-
False
-
None
-
False
-
-
-
-
-
-
-
Clustering: performance degradation in ejb-servlet scenario
In the EJB+SERVLET scenario (4 nodes cluster using both web and ejb replicated cache with MOD_JK load balancer) we noticed a ~10% in performance degratation between EAP7 and EAP8:
- average response time EAP-8.0.0.Beta-CR1: ~6.83 msec (see and )
- average response time EAP-7.4.8.GA-CR2: ~5.22 msec (see )
- average response time EAP-7.4.7: ~5.12 msec
Cluster nodes are configured as follows:
embed-server --server-config=standalone-ha.xml if (outcome != success) of /subsystem=jgroups:read-attribute(name=default-stack) /subsystem=jgroups/channel=ee:write-attribute(name=stack,value=tcp) else /subsystem=jgroups:write-attribute(name=default-stack,value=tcp) /subsystem=jgroups/channel=ee:write-attribute(name=stack,value=tcp) end-if /subsystem=infinispan/cache-container=web/replicated-cache=testReplWeb:add() /subsystem=infinispan/cache-container=web/replicated-cache=testReplWeb/component=locking:write-attribute(name=isolation, value=REPEATABLE_READ) /subsystem=infinispan/cache-container=web/replicated-cache=testReplWeb/component=transaction:write-attribute(name=mode, value=BATCH) /subsystem=infinispan/cache-container=web/replicated-cache=testReplWeb/store=file:add() /subsystem=infinispan/cache-container=web:write-attribute(name=default-cache, value=testReplWeb) /subsystem=infinispan/cache-container=ejb/replicated-cache=testReplEjb:add() /subsystem=infinispan/cache-container=ejb/replicated-cache=testReplEjb/component=locking:write-attribute(name=isolation, value=REPEATABLE_READ) /subsystem=infinispan/cache-container=ejb/replicated-cache=testReplEjb/component=transaction:write-attribute(name=mode, value=BATCH) /subsystem=infinispan/cache-container=ejb/replicated-cache=testReplEjb/store=file:add() /subsystem=infinispan/cache-container=ejb:write-attribute(name=default-cache, value=testReplEjb)
No exceptions are thrown in the server logs which might look as the culprit for the degradation;
Looking at the graphs it looks like the situation gets worse over time (tests last about 18 minutes and degradation is worse in the last half);
- is caused by
-
JBEAP-24258 UT005023: Exception handling request to /EjbCounter: java.util.NoSuchElementException
- Closed
-
WFLY-18065 Distributed @SessionScoped @Stateful EJBs require excessive cache transactions per invocation
- Closed
-
WFLY-18036 Marshalling optimizations are not getting applied to @SessionScoped @Stateful EJBs
- Closed
- is related to
-
WFLY-18023 @SessionScoped EJBs are replicating proxy placeholders unnecessarily
- Closed