SAVACT, checkpointing and tape movment

["Charles Wilt" <cwilt@xxxxxxxxxxxx>]

All,

 

Interesting problem, I've got a SAVLIB command saving two libraries with the
following SAVACT parms:

savact(*SYNCLIB) savactwait(5)

 

Recently, started having problems where the locks by the save job are
interfering with a extranet web application.

 

It doesn't happen all the time.  Looking at the SWA messages on a night
without problems I see:

17:06:14 - Save starts

17:06:43 - All Objects locked in LIB1

17:06:46 - All objects locked in LIB2, checkpoint processing in progress

17:10:48 - Checkpoint Processing complete

 

One a night with problems I see:

17:09:15 - Save starts

17:09:35 - All Objects locked in LIB1

17:09:36 - All objects locked in LIB2, checkpoint processing in progress

17:55:25 - Checkpoint Processing complete

 

Now as it turns out, on the nights that have had a problem, the wrong tape
was being used.  Instead of the current save going at the end (sequence #16)
of a daily tape, the save was going to the end (sequence #202) of a full
system save tape.

 

If I'm reading the v5r3 material I have correctly, the physical tape's not
supposed to start moving till the checkpoint has been reached. But this is a
v5r4 box, perhaps something has changed?

 

Can anyone confirm or deny that SWA checkpointing would be delayed by the
need to position the tape to the *END at v5r4?

 

If the extra tape movement is not the problem, I'm at a lost as to why
checkpointing might take 45 minutes when it usually only takes 4.  These
aren't big libraries, one has 501 objects the other has 81.  I see no
messages about the SWA timing out, or objects not being saved, even with the
SAVACTWAIT of only 5 seconds.

 

One last piece of additional information, on the save report all the logical
files in both libraries have a "Save While Active Timestamp" of 17:09 while
the physical files all have a "Save While Active Timestamp" of 17:55 on a
night with the problem.

 

 

Thanks!

Charles Wilt