Jobs stay in status starting

[faris_3]

Hi,

Did you try moving the queue manager on another node ? (sometimes this solves this type of problem).

You should also check that you use the latest IPC patch.

/homi

[Wim Van den Wyngaert]

Home : no. Queue manager wan't listening.

And no IPC patch (policy is not to install patches after production started).

Wim

[Volker Halle]

Wim,

what's the state of JOB_CONTROL in the forced dump ? HIB or LEF ?

JOB_CONTROL has 2 internal work-queues. There once was a problem, which may have caused problems with those queues on Alpha. It was apparently solved past-V7.3.

What's the image ident of your JBC$JOB_CONTROL ?

Volker.

[Wim Van den Wyngaert]

State HIB, ident X-27 (dd 19-dec-2002)

Wim

[Thomas Ritter]

Wim, are the batch queues set up for automatic failover or autostart enabled. Not sure on the terminolgy.
List the output for the some queues with
$ show queue/full


typical is

Batch queue GENERIC_BATCH, idle, on NODEA::
/AUTOSTART_ON=(NODEA::,NODEB::,NODEC::) /BASE_PRIORITY=3 /JOB_LIMIT=1 /OWNER=[SYSTEM] /PROTECTION=(S:M,O:D,G:R,W:S)

[Thomas Ritter]

forgot to ask please show the listing from

$ sh que/manager/full
Master file: COMMON:[SYSEXE]QMAN$MASTER.DAT;

Queue manager SYS$QUEUE_MANAGER, running, on NODEA::
/ON=(*)
Database location: COMMON:[SYSEXE]

[Wim Van den Wyngaert]

Thomas,

Master file: disk$salp_conf:[QUEUES]QMAN$MASTER.DAT;

Queue manager SYS$QUEUE_MANAGER, running, on SALPV1::
/ON=(SALPV1,SALPV2)
Database location: disk$salp_conf:[QUEUES]


Batch queue SALPV1$BATCH, idle, on SALPV1::

/AUTOSTART_ON=(SALPV1::) /BASE_PRIORITY=4 /JOB_LIMIT=3 /OWNER=[SYSMGR,SYSTEM] /PROTECTION=(S:M,O:D,G:R,W:S) /RETAIN=ERROR

Wim

[Thomas Ritter]

Wim, I took an interest in this problem thinking that with our 4 node DT 7.3-2 cluster and diverse work mix that if we don't experience the problem, then maybe problem is setup or resource related. We can report with out 400 or so queue and 1000s of daily batch jobs that all is well. We requlary failover the queue manger from one node to another.
We do use the /on=(*) qualifier on the queue manager startup.
Maybe next time using AMDS or SDA try to exclude some I/O sub-system related delay. Maybe as others have written, the accountng resource or other were not available. Did OPCOM report any problems ?

[Wim Van den Wyngaert]

Thomas,

I permanently monitor everything I can think of (all process quota & channels, all memory things, all abnormal queue things, abnormal process things ...).

OPCOM reported nothing abnormal.

Note that this is a 2 node server cluster with 9 pagefile-only stations as cluster memebers (but not queue manager runners) and it's the first time in 5 years we have the problem. And only on this cluster and only once.

Next time I do an autogen before I reboot ...

Wim