The SARPCD process occasionally fails to start during startup after a Fix Pack (PTF) installation/uninstallation

Content

Cause:
A failure to start the SARPCD process (RPC daemon) happens when starting a CICS region, SFS server or while starting SARPCD process after installing/uninstalling the Fix Pack (PTF). A possible cause is due to the shared memory segments and semaphore resources that are not cleaned up by the SARPCD process when it shuts down. These segments are owned by TXSeries and
used for CIPC communication. When starting, the SARPCD process first attempts to clean these shared memory segments and semaphore resources which were created by the previous level of TXSeries. A failure might occur during this cleanup activity. The failure may show as either of the two different symptoms:

1. The first symptom is message "ERZ058504I/0107: Starting RPC daemon" is presented and followed by a hang condition.
2. The second symptom is a message returned when starting the RPC daemon indicating it could not be started. Messages displayed are:

ERZ058504I/0107: Starting RPC daemon.
ERZ058503E RPC daemon could not be started.

Resolving the problem:
The solution is to check and clean up the shared memory segments and semaphore resources manually that exist for CICS. The stopped SARPCD process implies that all CICS regions and SFS servers are stopped on the system. Check for any running CICS processes and stop all of the processes before performing the clean up.

Steps to clean up the shared memory segments:

1). Check for existing segments by issuing the command:
ipcs -a |grep cics

When shared memory segments and semaphore resources are present, the information returned will be similar to:
m 1409031 0xeb200002 --rw-rw---- cics cics root sys 0 26214400 252
3722 13:58:33
s 8220 0x00000000 --ra-ra---- cics cics root sys 2000 13:59:34 14:10:21

The output shows two segments present. The first is a shared memory segment (the line starts with"m"). The second is information related to a semaphore resource (the line starts with "s").
The ID of the memory segment is needed for the next commands. The ID is in the second column. On Solaris, the numbers are smaller, made up of one or two digits. On HP and AIX the number are larger.

2). Clean up the shared memory segments with command:
ipcrm -m <id>

3). Clean up the semaphore resources with command:
ipcrm -s <id>