Setting SafetyLevel=normal can cause cicsas failure in scenarios where a library or application loaded in the cicsas executes shmat call with shmaddr set to NULL.
When you set the SafetyLevel to normal in the RD.stanza, the region attaches and detaches itself at the address specified in RegionPoolBase attribute of RD.stanza with every cics api call. If shmat is used in the application program or in other products (like DB2 or MQ or any other product whose library gets loaded into the cicsas memory) which is being used in conjunction with TXSeries for attaching to the shared memory and passes NULL as a second argument (shmaddr) for shmat, the shared memory is attached at the first available address as selected by the Operating System.
It is possible that the Operating System can assign the same address as RegionPoolBase address. So, the next time when the TXSeries region tries to attach to the address specified in RD.stanza for RegionPoolBase, it fails with EINVAL.
The following message gets recorded in symrecs file:
SYMPTOMS = TIME/"09/22/10 12:37:08.927364556" REGION/TEST PROD/5724AX710 LVLS/710 MOD/"@(#)conco, 17:23:35, Jul 27 2010" FUNC/ConCO_WaitForAnyAMChild LINE/676 MS/010089 MSN/367 SRC/2 PRCS/0 ABCODE/ SRVID/5 PID/1437874 TID/7 PROC/cicsam
SECONDARY SYMPTOMS = Child(PID 250074) returned with exit value 6
Set SafetyLevel=none in this scenario, which will make the cicsas to remain attached to the region pool throughout the lifetime of the cicsas.