IBM BPM cluster fails to start

Question

When I started the Process Center from the Process Center Quick Start console, the deployment manager profiles starts, the node starts but the cluster fails to start. What could be the reason? I do not see any errors in the log. Here's the screenshot of where it halts:- 

Answer 1

"Deployment manager starts, node agent starts, cluster fails" on a Process Center almost always has one of these causes - in the order to check:

1. Database: the cluster members need BPMDB / PDWDB / CMNDB; if the database is down, credentials expired, or the JDBC port is blocked, the member stops during BPM initialisation. Look in the member's SystemOut.log for DSRA / CWLLG errors and J2CA0056I; test with WebSphere admin console > Data sources > Test connection.
2. Messaging engine: the SIBus messaging engine must start before the BPM applications; a locked file store (CWSIS1519E - another JVM holds the lock, or the file store is on a path the member cannot write) or a full disk stops it. Check <profile>/logs/<member>/SystemOut.log for CWSIS / CWSID messages.
3. Ports / hostnames: after a hostname or IP change (VM cloning, DHCP) the member cannot bind or cannot reach the deployment manager; WSVR0009E / "port in use" or ADMU3011E. Verify serverindex.xml hosts and netstat.
4. Node not synchronised: a configuration change was saved but the node was not synchronised - syncNode.sh dmgrhost 8879 -username ... -password ... with the node agent stopped, then start again.
5. Memory: the member starts and dies with OutOfMemoryError on a small box (Quick Start VMs): raise the heap in the server's JVM settings or free memory.
6. Timeout of the Quick Start console: the console gives up while the cluster is still starting; check whether the member actually came up (serverStatus.sh -all) before assuming failure.

# start manually to see the real error
<profile>/bin/startServer.sh SingleClusterMember1            # instead of the Quick Start console
tail -f <profile>/logs/SingleClusterMember1/SystemOut.log    # first "E" line after "Starting server" tells the cause
grep -E " E |OutOfMemory|CWSIS|DSRA" <profile>/logs/SingleClusterMember1/SystemOut.log | head

The startServer.log and SystemErr.log of the member complete the picture; ffdc/ holds the first-failure data. If the member starts manually but not from the console, the console's timeout is the only problem.

References

• BAW documentation - starting and stopping servers, troubleshooting startup