You say that the problem happens when all the tasks terminate. Your problem is with not enough LQSA for termination. During termination a number of RBs are getmained by RTM to handle termination - like an RB that your ESTAE gets control under (a PRB, IIRC). Or a PURGEDQ SVRB. Depending on what your ESTAE does, you'll need more LSQA for further stuff.
I don't have a rule of thumb how much LSQA is needed per TCB. Given that you say you create 1000 tcbs, and each tcb creates at least one subtask, we're talking at least 2000 TCBs. Plus their associated RBs and CDEs. I'd guess that you need at least 6MB below the line of storage reserved for LSQA, possibly more. The only way to do that is to write a custom IEFUSI that really reserves that much LSQA especially for your job. Or the equivalent parmlib member. Remember that LSQA 'grows' from top of region below downwards while private storage 'grows' from bottom of region upwards. So conditional getmains don't help here IMHO. You would have to determine current top of region programmatically and then subtract 1-2MB for termination and then check if you've still got enough room to do your getmain. Anecdote: Before IBM introduced command classes and all the messages that go with too many commands issued at the same time there used to be regular wait states (wait07E, IIRC) due to LSQA exhausted in *master*. Commands execute in *master* (for the most part). Too many commands at the same time generated the exact same situation you currently have - not enough LSQA left. Which is really deadly when it happens in asid 1. IBM only allows 100 commands per class these days. If more are issued, they get held back until there's 'room' again to have them execute. Why do you need 1000 tcbs? Regards, Barbara ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN
