Do you run CICS v5 on HP? What's your maxssiz set to?

[sjwales]

We run CICS/TX-Series 5 (CICS at PTF 4, Encina to PTF 3) on HP-UX 11.11 with Oracle 9.

In one of our regions, we keep getting core dumps in /var/cics_regions/<region> for that region. We've been wondering what's been causing them and I think I finally found a connection yesterday.

I was logged in to the shell yesterday as root and while su'ed to oracle doing some other work, I kept getting notifications that certain PID's were being killed because they were running out of user stack, and that possible causes were out of memory (no), out of swap space (no) or user stack size exceeded maxssiz. Each one of these messages gave me another core file.

As per HP recommendations for Oracle 9i, we have maxssiz set to 128MB. HP Documentation says that the default is 8MB, max setting 200MB. We have a reboot scheduled for the 12th and at that stage, we're going to set maxssiz to 160MB.

I was wondering is there anyone else running CICS 5 on HPUX 11i and if so if you'd be willing to share your value for this kernel parameter. (There are two, maxssiz and maxssiz_64, I'm only looking for maxssiz).

Thanks
Steve

 

[cgainey]

Steve,

It's been a while since you posted this message so I hope that you have resolved your issue. J-I-C, With the exception of running CICS PTF5 with Encina PTF3, our enviroments are identical. However, we did not have to make any modifications to the maxssiz kernel parameter. Here is the listing of the parameters that were suggested to be modified by our consultant. Hope this provides some assistance. Note that some of the parameters are dependant on others so you might not be able to modify them in quite the same order as they are listed.

maxfiles
4096

maxfiles_lim
4096

max_thread_proc
1024

maxuprc
512

nproc
1024

nflocks
8192

ninode
2048

nfile
4 * ninode value

msgseg
32767 (or less)

msgmnb
65 535

msgmax
65 535

msgtql
1024

msgmap
258

msgmni
256

msgssz
16 32 for ctg

semmni
512

semmap
514

semmns
1024

semmnu
1020 (nproc value minus 4)

shmmax
483 183 821
(Ensure that this parameter is set to 483 183 821 or 90% of the physical memory (in bytes), whichever is higher. For example, if you have 512 MB of physical memory in your system, set shmmax to 483 183 821 (512 * 0.9 * 1024 * 1024).


shmseg
16




~asante,<cag>

 

[sjwales]

Thanks for the followup. We upped our maxssiz last reboot anyway - seems to have reduced the aborts, but hasn't eliminated them.

As a curiosity, what is your maxssiz set to?

We have a couple of other parameters less than what yours are set to, I might have to do a little more research on them

Steve

 

[cgainey]

Steve,

My apologies, that was your original question. Our maxssize is set to '0x800000', which is calculated to '8388608'. Again we didn't have to modify this parameter for CICS or Oracle, this was our default setting. Just as an afterthought, have you considered upgrading CICS from PTF 3 to PTF 5. I know that there were a lot of fixes in between PTF's that may address your issue.

If you have a development server that you can test on before making the changes in production, you an travel to this link,

http://www-306.ibm.com/software/htp/txseries/support/

Click on 5.0 in the 'Downloads' section and you will find a list of the PTF that are available. There could be more available but if they are not on this page available for download then I wouldn't trust them.

Hope this helps.


~asante,<cag>

 

[cgainey]

Also Steve, is CICS logging any type of error message in the console.0000xx log when these core dumps happen?



~asante,<cag>

 

[sjwales]

We have PTF 4 at the moment with the ifix and have just received word from Mincom this morning that we need to apply PTF 6 to correct another problem that we're having in another region.

So that's on the slate here pretty darn quick

Also, the message is usually something like this:

Transaction 'CPMI', Abend 'A147', at '????'.

An A147 abend is fairly generic as far as I can tell. From the CICS documentation:

A147
Explanation: The transaction was purged.
System Action: CICS abnormally terminates the transaction.
User Response: This abend indicates that the transaction was purged, either via CEMT or as the result of an immediate region shutdown, or because an exceptional condition was detected. Resubmit the transaction.

Doesn't help much

Steve

 

[cgainey]

Yeah some of those error messages are not helpful at all for troubleshooting. Sorry that I could not have been of more assistance. It is interesting that you are having problems with your configuration and ours is working smoothly. When you get an opportunity please try the kernel parameters that I sent in an earlier post and let me know how they work. Those parameters are from a CICS document called 'Ellipse Install Preparation for HPUX 11i' supplied by our Mincom consultant.

~asante,<cag>