I have bpel process with web service (which only print number of loop cycle) deployed on cluster1 and cluster2. I'll send soap message to process on cluster1 and process is running in for loop. In 50th loop It will wait and You can kill cluster1 with signal 9. The cluster2 should takeover process, but after some ConnectExceptions It do NOTHING.
BTW I tried Riftsaw 2.3.1-Snapshot on SOA-P 5.2.ER1 with fixed issue https://issues.jboss.org/browse/RIFTSAW-404 and cluster2 tried to restore process, but It failed in every attempt with this exception:
2011-08-02 15:36:31,475 ERROR [org.apache.ode.scheduler.simple.SimpleScheduler] (ODEServer-21) Error while processing a persisted job: [JobId: hqejbhcnphr6hf75eilk7w,nodeId: 127.0.0.1:1199,scheduled: false,transacted: true,ts: 1312292188221,channel: null,instaceId : 0,type: INVOKE_CHECK,retrycount: 2]