Matt,

> V SMS,PDSE,ANALYSIS
> IGW031I PDSE ANALYSIS  Start of Report(SMSPDSE ) 879
> ++ Unable to latch ASRBULCH:000000007FF94DC0
>    Latch:000000007FF94DD8 Holder(006A:00AFF208)
>    Holding Started Task:TECHCICS
> ---------data set name---------------------- -----vsgt-------
> SYSVIEW.R137.CNM4BLOD.SYS1                   01-SYS002-000614
> ++ Unable to latch HL1b:000000007FF67600
>    Latch:000000007FF67610 Holder(006A:00AFF208) IGWLHPRG
>    Holding Started Task:TECHCICS

Good old ASRBULCH latch. First of all, make sure you have all the required 
maintenance for PDSE on (there are several apars involving ASRBULCH out there, 
and this appears to be an ongoing bug that IBM seems unable to fix). 

> VARY SMS,PDSE1,FREELATCH(000000007FF67610,006A,00AFF208),RETRIES(5)
> IGW032I PDSE FREELATCH  Start of Report 889
> ++ Free latch failed latch:7FF94DD8 is either broken or not a latch
> PDSE FREELATCH  End of Report

In my case (we had it twice) it was not obvious which latch to free (there were 
tons of ASRBULCH ones). So I went through all of them and eventually found the 
one that was left and had caused everyone else to back up. In that process, I 
also generated some PDSE abends complaining about freed latches. 

By the way, even forcing the address spaces will not clear the contention!

Admittedly, I have never seen the response of 'either broken or not a latch'. 
That sounds like something went ahead and freed the storage where the latch is 
situated.

I have heard from our customers (who were also affected for a library that is 
NOT in linklist) that restarting SMSPDSE1 did not free the latches. The 
customers affected by it were also able (with trial and error) to use freelatch.

If that library SYSVIEW.R137.CNM4BLOD.SYS1 is not a load library (or anthing 
else that *requires* it to be a PDSE) then I suggest to make it a PDS to avoid 
hitting PDSE code.

If you have the appropriate contract with IBM, when this hits next, then take a 
dump of the address spaces involved in the latch contention and SMSPDSE/1 and 
report the bug to IBM.

Regards, Barbara Nitz

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to