On Wed, 10 Jun 2009 15:06:54 -0500, Dave Butts <[email protected]> wrote:
>It was an OS problem after all, and not due to the MCLs.
>APAR OA29370 has been opened.
>
>We have been over the 65535 device boundry for a long time, but the bug
>became exposed when I deleted over 5000 addresses the week before via
>dynamic iogen activate. The PAV assignment code broke because of the 2
>byte CMCTHMBI field. We have just been lucky for several years that the
>truncated value (to fit the field) resulted in a value that was higher than our
>highest UCB. With the recent deletion of the 5000+ devices along with the
>truncation, we ended up with a total number of devices being tracked that is
>less than our starting DASD UCB.
>
>Strange though that the propblem didn't actually start until the next weekend
>when we performed a power-on-reset.
>Still investigating that one...
>
Here's another one....
APAR Identifier ...... OA29016 Last Changed ........ 09/06/25
UNABLE TO BIND PAV DEVICES FOLLOWING AN ACTIVATE FAILURE
Symptom ...... IN INCORROUT Status ........... CLOSED PER
Severity ................... 2 Date Closed ......... 09/06/25
Component .......... 5752SC1C3 Duplicate of ........
Reported Release ......... 730 Fixed Release ............ 999
Component Name IOS Special Notice HIPER
Current Target Date ..09/08/28 Flags
SCP ................... FUNCTIONLOSS
Platform ............
Status Detail: APARCLOSURE - APAR is being closed.
PE PTF List:
PTF List:
Release 730 : PTF not available yet
Release 730 : Relief is available in the form of: PTF
Release 740 : PTF not available yet
Release 740 : Relief is available in the form of: PTF
Release 750 : PTF not available yet
Release 750 : Relief is available in the form of: PTF
Release 760 : PTF not available yet
Release 760 : Relief is available in the form of: PTF
Parent APAR:
Child APAR list:
ERROR DESCRIPTION:
Customer performed an ACTIVATE SOFT, which involved changes on
the usage of base and alias devices. During dynamic
configuration change processing, IOS sets indicators in the
devices being changed to prevent them from being bound again
during the dynamic change. However, if the activate fails
and all traditional PAV devices were unbound, configuration
change processing fails to reset these indicators.
This prevents these devices from future binding.
VERIFICATION STEPS:
1. Binds are not occurring following an ACTIVATE TEST
2. Check SYSLOG for IOS500I message to determine if TEST is
used with ACTIVATE.
3. Get a dump of IOSAS very soon after failure (or if can
recreate, after you recreate)
4. Run CTRACE COMP(SYSIOS) SUMM LOCAL
5. Find the date in question F 'mm/dd/yyyy'
6. F 'Dyn config change - Start' to find the start of the
activate.
7. F CMV1 eyecatcher entries after start for one device in
question.
8. Check the UCBXNBND Counter for that device and then find the
device again checking to see if counter is non-zero.
UCBXNBND bit is UCBX+6F = 80 bit.
Counter is UCBOB + x'34'
LOCAL FIX:
PROBLEM SUMMARY:
****************************************************************
* USERS AFFECTED: Users at HBB7730 and above. *
****************************************************************
* PROBLEM DESCRIPTION: After a failed ACTIVATE, unbound *
* traditional PAVs are unable to bind or *
* perform I/O because the back out *
* processing is incorrectly skipped for *
* these devices. *
* PAV PAVS D/T2105 D/T2107 D/T1750 *
****************************************************************
* RECOMMENDATION: *
****************************************************************
During the course of an ACTIVATE, PAV Aliases undergo
processing to prevent them from being bound while dynamic
changes to the configuration are happening.
If an error occurs during an ACTIVATE, this processing should
be backed out so that the Aliases can be bound as before.
However, if only PAV Aliases of the unbound traditional type
are being changed, the back out processing does not occur as
expected and attempts to bind the effected PAVs will fail.
PROBLEM CONCLUSION:
Error processing will be changed so that all PAV Aliases will
undergo back out processing to be returned to their original
state and therefore be able to be bound.
--
Mark Zelden
Sr. Software and Systems Architect - z/OS Team Lead
Zurich North America / Farmers Insurance Group - ZFUS G-ITO
mailto:[email protected]
z/OS Systems Programming expert at http://expertanswercenter.techtarget.com/
Mark's MVS Utilities: http://home.flash.net/~mzelden/mvsutil.html
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
