View previous topic :: View next topic
|
Author |
Message |
BernardMein
New User
Joined: 12 Nov 2008 Posts: 5 Location: Springfield MA
|
|
|
|
Does anyone know why the SMF datasets would be swapped after an IPL?
We use the standard of two SMF datasets per LPAR and have automated most of the SMF processing (ie dumping the datasets when they're full).
Let's say an LPAR is about to be brought down and IPLed, and one SMF dataset has a state of ALTERNATE and the other is ACTIVE at 1% full.
Let's say that the result of the system being brought down and the subsequent IPL creates 9% more SMF data. After the IPL, the state of the SMF datasets now shows one as ACTIVE and the another as DUMP REQUIRED. The one with DUMP REQUIRED is only 10% full.
Why does SMF or MVS close the SMF dataset that is only 10% full? I don't know if the closing of the SMF dataset occurs when the system is brought down or brought back up.
Thanks |
|
Back to top |
|
|
Bill Dennis
Active Member
Joined: 17 Aug 2007 Posts: 562 Location: Iowa, USA
|
|
|
|
I've noticed that at IPL the system always switched to the first SMF dataset. Never saw it documented as to why. |
|
Back to top |
|
|
nevilh
Active User
Joined: 01 Sep 2006 Posts: 262
|
|
|
|
when the system is stopped normally the operators will do a Z EOD this will write the smf buffers and issue a close amongst other things |
|
Back to top |
|
|
Robert Sample
Global Moderator
Joined: 06 Jun 2008 Posts: 8696 Location: Dubuque, Iowa, USA
|
|
|
|
I'm not finding any documentation on it, but I did an IPL yesterday and the system went through and verified each SMF dataset in reverse order. Once the verifies were done, the first file started being used for recording. I suspect the system automatically starts with the first SMF dataset no matter which was being used before the IPL. |
|
Back to top |
|
|
BernardMein
New User
Joined: 12 Nov 2008 Posts: 5 Location: Springfield MA
|
|
|
|
Thank you everyone,
I checked the procedures of our Operations staff and sure enough the "Z EOD" command is there. The IBM documentation for this HALT command isn't clear about the use of the first SMF dataset, as one sentence indicates always switching to the first dataset and another indicates switching to the ALTERNATE dataset; I'll check the logs following the IPLs over the weekend. It's nice to have a cause (and understanding) associated with the SMF state at IPL.
Thanks. |
|
Back to top |
|
|
Bill Dennis
Active Member
Joined: 17 Aug 2007 Posts: 562 Location: Iowa, USA
|
|
|
|
We rarely went throught the formal shutdown of "Z EOD" but our system always went back to the first dataset anyway. |
|
Back to top |
|
|
nevilh
Active User
Joined: 01 Sep 2006 Posts: 262
|
|
|
|
The following is a taken from the SYSTEM COMMANDS manual
Quote: |
The information gathered by SMF and user-written exit routines is recorded
on direct access volumes in one of the SMF data sets. These data sets,
called primary and secondary data sets, must be online at system
initialization. At that time, SMF uses the primary data set as the active
recording data set unless it is full. If the primary data set is full, SMF
checks each data set in the order it is listed until it finds one that is
not full. SMF then uses this data set as the active recording data set and
requests that the operator dump all data sets that are not empty. |
Which seems to say that it will use the primary dataset unless it is full |
|
Back to top |
|
|
Srinivasa Rao
New User
Joined: 21 Jun 2005 Posts: 75
|
|
|
|
SMF datasets are defined to collect statistical information. When active dataset is full at time of shutdown after issuing Z EOD the system will switch the datasets. This is done by an exit defined in the system. The switch is based on the message IEE362A using the command "I SMF".
But in our system we defined it in automation different from exit. |
|
Back to top |
|
|
enrico-sorichetti
Superior Member
Joined: 14 Mar 2007 Posts: 10873 Location: italy
|
|
|
|
we do know what SMF datasets are and the dataset full logic
the thread is about the dataset selection logic at IPL time
ROTATE mode, use the next one available
or use he first available
rotate would seem a better choice, is more forgiving of operation errors
if a dump is not taken at the proper time SMF data is not mixed up
( in other words an SMF dump will contain contiguous timestamps only ) |
|
Back to top |
|
|
|