Question & Answer
When initializing multiple CICS Transaction Server for z/OS (CICS TS) regions about the same time, the LLA service fails with abend S878 and many (but not all) of the CICS region fail to initialize. The reasons for the CICS failures can be quite different. For instance, some of the CICS failures have been:
DFHLD0001 An abend (code 0C4/AKEX) has occurred at offset X'FFFF' in module DFHLDLD1.
DFHLD0001 An abend (code 0D5/AKEX) has occurred at offset X'FFFF' in module DFHLDLD1.
DFHPC0401 Abend AEXZ issued by III task.
DFHLD0001 An abend (code 878/AKEX) has occurred at offset X'FFFF' in module DFHLDDMI
DFHLD0002 A severe error (code X'6929') has occurred in module DFHLDDM.
Using IPCS to look at a sytem dump of LLA at the time of the abend S878 (abendS878), the IPCS VERBEXIT VSMDATA command shows a buildup of subpool 229 key 0 (K0 zero ) as well as SP252. SP229 storage is filled with LLNH control blocks and the SP252 with LLXE control blocks. These are related to the number of LLACOPY requests coming in. The buildup of these control blocks suggests the requests are coming in at a rate higher than LLA can process.
Using IPCS' LLA CTRACE command IP CTRACE COMP(SYSLLA) FULL LOCAL shows several thousand requests in the last couple seconds. This is more than LLA can handle at one time, leading to the short on storage (SOS).
There are only two options to decrease the rate at which these requests come:
Change the CICS regions' system initialization table (SIT) parameter LLACOPY to specify something other than YES. Either NEWCOPY or NO will cause CICS to use the BLDL process instead of LLACOPY during initialization.
Stagger the startup of the CICS regions so that they do not all start at the same time.
CICS/TS CICSTS CICS TS CICS Transaction Server
18 January 2018