%QMAN-W-LOWMEMORY

[Volker Halle]

Art,

executing batch and print jobs continue unharmed, if you restart the QUEUE_MANAGER. On my standalone V8.3 test system, it took about 21 seconds to write the .DMP file - I at least verfified, that the command worked, before I posted it.

The time for writing the forced dump certainly depends on the virtual memory used by the QUEUE_MANAGER. Did you check the size and the creation/modification date of SYS$SYSTEM:QMAN$QUEUE_MANAGER.DMP ?

Volker.

[P Muralidhar Kini]

Hi Art,

>> Great, except these systems are v7.2-2 and aren't going to be upgraded.
Upgrading would have given you access to the patches which has the
fix for the problem.

As upgrading is not a option, other alternative is to use a workaround.
i.e. to increase the pagefile size. Again with this change also, the problem
might come back again. Its a tricky situation.

As of now, how frequently are you seeing the "QMAN-W-LOWMEMORY" messages?

Regards,
Murali

[Art Wiens]

Thanks Bob. Hadn't quite made it out of first gear yet ;-)

Volker, the dump file only took 45 seconds to write, but for some reason the queue manager was working "very hard" when it restarted. Definately "paused" the cluster for several minutes while it COMputed something.

Thanks,
Art

[Art Wiens]

Murali, since the queue manager reset 40 minutes ago, I have not seen the message again. The queue manager's pagefile quota and count are back to "normal":

$ write sys$output f$getjpi("2030CD04","PGFLQUOTA")
649984
$ write sys$output f$getjpi("2030CD04","PAGFILCNT")
592080

Cheers,
Art

[P Muralidhar Kini]

Hi Art,

>> since the queue manager reset 40 minutes ago, I have not seen the
>> message again. The queue manager's pagefile quota and count are
>> back to "normal"
Hmm. Looks like the workaround would be to restart the queue manager
at a silent time.
Increasing the pagefile quota might not be required as the problem may not be
related to that (i.e. problem is due to overflow of longword in the code)

Regards,
Murali

[Volker Halle]

Art,

you're NOT suffering from the bogus QMAN-I-LOWMEMORY problem addressed in those patches !

You seem to have a REAL virtual memory 'leak' in your QUEUE_MANAGER process.

The LOWMEMORY situation will be reported, if
PAGFILCNT (remaining pagefile quota) is less than 10% of PAGFILQUO (pagefile quota at start of process = same as for JOB_CONTROL).

The JIB data you've posted confirms this problem:

JIB$L_PGFLQUOTA 00009EB0
JIB$L_PGFLCNT 00000950

2384 < 10% of 40624

Based on the uptime of your system (328 days), this may be a slowly building memory leak, if we assume, that the QUEUE_MANAGER process had about the same uptime.

Maybe make a note in your calendar for 28-MAR-2011 to restart QUEUE_MANAGER, before you see this problem again...

Volker.

[Volker Halle]

Art,

QUEUE_MANAGER inherits it's pagefile quota from JOB_CONTROL. JOB_CONTROL is being started with a pagefile quota of 1/2 of the total available page file space in the system at the time JOB_CONTROL gets started (see SYS$STARTUP:VMS$CONFIG-050_JOBCTL.COM).

Your page files aren't 'large' enough to trigger the math overflow problem, so you must be experiencing a real LOWMEMORY problem.

Consider to watch PAGFILCNT of the QUEUE_MANAGER process over time. You'll start to get those %QMAN-W-LOWMEMORY messages, if remaining pagefile quota gets near 10% of the initial pagefile quota value for the QUEUE_MANAGER process.

Volker.