is there anybody experienced in tuning memory pools in CICS running JAVA transactions?. We are unsure about the number of JVMs we could safely run.
Also in certain situations cics began restarting JVMs very frequently, we got many J9 TCB mismatches and cics consumes enormous CPU. Number of tasks is raising until eventually DSA memory is exhausted and CICS stops responding.
I can provide the CISC statistics if there is somebody to have a look.
Joined: 14 Jan 2008 Posts: 2504 Location: Atlanta, Georgia, USA
When you say "DSA" are you referring to 24-Bit User Storage or 31-Bit, which is "EUDSA"?
In the SIT, what is your EDSALIM parameter set to?
If you're not sure, issue a CEMT INQ SYSTEM and review the DSALIMIT and the EDSALIMIT.
Hopefully, the majority of your application programs run in Key 9 (EXECKey : User) and DAtalocation : Any (program is linked as AMODE 31 RMODE ANY) or perhaps 64-Bit (above the bar), which Java supports.
Sounds like you're running Threadsafe, having these multiple Java TCB's?
If you are running Threadsafe, ensure the applicable PPT entries specify COncurrency : Threadsafe and meet compliance for Threadsafe.
If any GLUE's or TRUE's are being used, ensure they are Threadsafe compliant as well.
What is your CICS Version/Release?
IMHO, IBM has sold the Mainframe community false promises because I hear nothing but problems related to Java and CICS.