3617610 - SAP instance goes down shortly after having being restarted by SUM

An SAP instance goes down shortly after having being restarted by SUM (e.g., during the "Restart System ..." step).

This causes SUM to become stuck at this phase.

At the trace files of the affected instance, errors related to semaphores being removed are observed just a few seconds after the instance was started.

For example, from a "dev_icm" trace:

[Thr ...] [date and time]
[Thr ...] *** ERROR => e=43 semop(131122,(36,0,0),(36,1,0),2) (43: Identifier removed) [evtux.c      1077]
[Thr ...] *** ERROR => EvtWtRst(17197) failed: 1 [mpixx.c      6350]
[Thr ...] *** ERROR => IcmJSessCommThread: MpiSelSelect failed(rc=3) [icxxjcom.c   3013]
[Thr ...] *** ERROR => e=22 semop(131122,(36,0,0),(36,1,0),2) (22: Invalid argument) [evtux.c      1077]
[Thr ...] *** ERROR => EvtWtRst(17197) failed: 1 [mpixx.c      6350]

When checking the SAP Start Service command line, the "-cleanipc XX" arguments are observed ("XX" stands for the SAP instance number).

For example:

sidadm   PID  ......   /usr/sap/SID/Instance/exe/sapstartsrv pf=/usr/sap/SID/SYS/profile/SID_InstanceName_hostname -cleanipc XX

• SAP NetWeaver based product
• ABAP Platform based product
• SAP running on Linux

KBA , BC-CST-STS , Startup Service , BC-UPG-TLS-TLJ , Upgrade tools for Java , BC-UPG-TLS-TLA , Upgrade tools for ABAP , Problem