Description of problem: The EJB client should do a retry if the invocation can not be processed because it reached the node in a state where the EJB is not available. This might happen during start/deploy/undeploy/shutdown. How reproducible: Due to another issue BZ900999 this might happen more often. I use a simple application and script to call the EJB in a loop during permanent server restarts or deployments. The problem happen with bz900999 in about 1 of 20 server stops Steps to Reproduce: 1. I've used the ejb-multi-server quickstart [1], use the test EJB of app-one or app-two deployed at two nodes. 2. The client application call the EJB in a endles loop. 3. The server is started/stopped by CLI with script: while [ 1 -eq 1 ]; do bin/jboss-cli.sh --connect --command="/host=master/server-config=app-oneA:stop(blocking=true)" bin/jboss-cli.sh --connect --command="/host=master/server-config=app-oneA:start(blocking=true)" done [1] https://github.com/wfink/jboss-as-quickstart/tree/ejb-multi-server/ejb-multi-server
Update - I've a solution to this which is currently being tested before I send out a PR.
Skynet must have started its career as a bugbot.
jaikiran pai <jpai> made a comment on jira WFLY-1308 Pull request sent
Pull request has been merged
Paul Ferraro <paul.ferraro> made a comment on jira WFLY-1306 Fix is just a workaround. Need to discuss a better way to improve transaction commit reliability with infinispan team.
Details inside vpn: https://post-office.corp.redhat.com/mailman/private/eap-pm-list/2013-July/msg00105.html