We recently had an occasion where it turned out the culprit was CICS system logging - or
rather deleting old achived systems logs. Seems when the retained log count is reached,
the system logger will delete the oldest archived log - before the switch takes place and the
CICS sync happens.
Someone decided to reduce the retained log count by many, many thousands.
James Campbell
On 30 Oct 2020 at 19:48, Donna Domovic via Internation wrote:
> Hello,
>
> Have a strange one that has us banging our heads on.
>
> We've noticed that we periodically have times where we see the Sync point dispatch time in the CICS monitor (ie SYSVIEW) showing large values (5+ seconds). Looking into it, the majority of this corresponds to large RMIDb2 times that aren't accounted for at the DB2 statement level. From what we've seen, the CPU does not appear to be maxed out during these periods.
>
> We investigated this using various monitoring tools such as CA Sysview/Insight for DB2, BMC Apptune,, and SMF reports but don't see any place on the DB2 side that accounts for this. All statements are executing quickly.
>
> As we've investigated this, we've also noticed that this appears to be isolated to 1 LPAR but not to a single CICS region, DB2 data sharing group, or application.. When it happens, it seems to impact all of them on that LPAR for the short period when it happens (less than 1 minute) which makes me lean towards it being something at the LPAR/system level rather than a DB2, CICS, or application problem.
>
> We are looking to see if we can find any pattern to when it happens and if there could be something running during that time that could account for this but so far have not had any luck despite different teams taking a look. Given the wide community here, I was curious if anyone has ideas or has seen something like this. If you have ideas on what we may be missing or what we might want to look at, I'm interested in those ideas as well.
>
> Thanks in advance for any information or ideas you provide.
>
> ------------------------------
> Donna Domovic
--
This email has been checked for viruses by AVG.
https://www.avg.com
Original Message:
Sent: 10/30/2020 3:49:00 PM
From: Donna Domovic
Subject: Cause of CICS Sync point dispatch time attributed to DB2
Hello,
Have a strange one that has us banging our heads on.
We've noticed that we periodically have times where we see the Sync point dispatch time in the CICS monitor (ie SYSVIEW) showing large values (5+ seconds). Looking into it, the majority of this corresponds to large RMIDb2 times that aren't accounted for at the DB2 statement level. From what we've seen, the CPU does not appear to be maxed out during these periods.
We investigated this using various monitoring tools such as CA Sysview/Insight for DB2, BMC Apptune,, and SMF reports but don't see any place on the DB2 side that accounts for this. All statements are executing quickly.
As we've investigated this, we've also noticed that this appears to be isolated to 1 LPAR but not to a single CICS region, DB2 data sharing group, or application.. When it happens, it seems to impact all of them on that LPAR for the short period when it happens (less than 1 minute) which makes me lean towards it being something at the LPAR/system level rather than a DB2, CICS, or application problem.
We are looking to see if we can find any pattern to when it happens and if there could be something running during that time that could account for this but so far have not had any luck despite different teams taking a look. Given the wide community here, I was curious if anyone has ideas or has seen something like this. If you have ideas on what we may be missing or what we might want to look at, I'm interested in those ideas as well.
Thanks in advance for any information or ideas you provide.
------------------------------
Donna Domovic
Exelon
------------------------------