EPICURE Software Release Note 122.0
When the QUEUE file is corrupted
Debra S. Baddorf
Most, if not all, of this information will be obsolete when we go to VMS V5.5. Bear this in mind!
During startup, there will be complaints the the queues are not running, the queue manager is not started or cannot be started, and/or that the queue file is corrupted. The name of the queue file is SYS$SYSTEM:JBCSYSQUE.DAT before VMS V5.5.
If you log on, when you do SHOW QUEUE you will be told that the queue manager isn't running. You will not be shown any queues. If you try to remedy this by logging in as SYSMANAGER and doing @CMN$STARTUP:STARTQUEUES, you will be told that the queue manager cannot start due to a corrupt queue file.
All of the following should be done as SYSMANAGER.
If the queue manager is still running on at least one node, log into
that node and go to step
. Until you STOP the queue
manager (or the node reboots) it may continue to run on the corrupt
queue file, so you may still have some nodes whose batch jobs are OK.
To find out if any nodes still have a running queue manager, type
$ SYSMANIf all nodes say that the queue manager is not running, then continue with this step, and try to force one to run.SYSMAN> SET ENV/CLUSTER
SYSMAN> SET TIMEOUT 0:2
SYSMAN> DO SHOW QUEUE
SYSMAN> EXIT
If all the queue managers are
stopped, then you have to force one to start. If this doesn't work,
then you'll have to abandon the attempt to ``recover the old queue file''
and go to the ``make a new queue file'' section
.
This trick is
``unsupported and undocumented'' but it works well enough for someone
else to have written these procedures, and it did work the only time
I tried it.
Pick a node and log into it (it might be safer not to use DAFFY or ELMER, unless those are the only nodes booted and you have a minimum cluster). Do the following to set the dynamic SYSGEN parameter JOBCTLD to 1 (this tells the queue manager to skip the file checking).
$ SYSGENIf the START/QUEUE/MANAGER part worked ok, you can now do SHOW QUEUE (on this node only). If it didn't work, then it gave an error message, and you'll have to go to the ``make a new queue file'' sectionSYSGEN> USE ACTIVE
SYSGEN> SHOW JOBCTLD
SYSGEN> SET JOBCTLD 1
SYSGEN> WRITE ACTIVE
SYSGEN> EXIT
$ START/QUEUE/MANAGER
$ SYSGEN
SYSGEN> USE ACTIVE
SYSGEN> SET JOBCTLD 0
SYSGEN> WRITE ACTIVE
SYSGEN> EXIT
You should already be logged into a node which has a queue manager running. Do the following:
$ SET DEFAULT USR$DISK1:[BADDORF.WIP.FIXQUE]This will do SHOW QUEUE commands with the output sent to files called FIXQUE_CHARS.LIST, FIXQUE_FORMS.LIST, and FIXQUE_QUEUE.LIST. It then parses the contents of these files to create a command procedure called FIXQUE_RELOAD.COM.$ @FIXQUE
Now, start a new, empty queue file and we'll put this save information into it.
$ STOP/QUEUE/MANAGERThis should reconstruct the information in the queue file. Proceed to the ``Post Cure'' section$ START/QUEUE/MANAGER/NEW MSYS_COMMON:[SYSEXE]
$ @FIXQUE_RELOAD
If you can't recover the old queue file, you need to create a new queue file from scratch. Do the following.
$ START/QUEUE/MANAGER/NEW MSYS_COMMON:[SYSEXE]This will reconstruct the bare queue files, but none of the jobs in the queues. They are lost. Proceed to the ``Post Cure'' section$ @SYS$MGR_UTIL:RESTORE_PRINT_FORMS
$ SYSMAN
SYSMAN> SET ENV/CLUSTER
SYSMAN> SET TIMEOUT 0:5
SYSMAN> @CMN$STARTUP:STARTQUEUES
SYSMAN> EXIT
If this queue file failure occurred during a cluster reboot (which it frequently does on a power glitch during backup jobs), then we need to go back and run all the stuff which should have run in batch queue during startup. It didn't run during the reboots since the batch queues were busted, remember?
First, create a temporary file in USR$SCRATCH:[SYSMGR]AFILE.TMP, and put the following lines in it:
$ @CMN$STARTUP:EPICURE_SYSSTARTUP$ scsnode == f$getsyi("NODENAME")
$ SYS_SUBMIT := SUBMIT /NOPRINT /QUEUE=`scsnode'_SYSTEM -
/USER=SYSMANAGER /LOG=SYS$MANAGER: /PRIORITY=255
$ SYS_SUBMIT /NAME=BOOTJOB SYS_STARTUP:BOOTBATCH.JOB
Now, do the following. Yes, if you came here from section
you are doing STARTQUEUES twice. That is on purpose, since some queues
won't be defined properly till certain other queues are created.
$ SYSMANYour temporary command file will run the epicure startup job and the other batch-based startup jobs which didn't happen during the reboot, due to the file problem.SYSMAN> SET ENV/CLUSTER
SYSMAN> SET TIMEOUT 0:5
SYSMAN> @CMN$STARTUP:STARTQUEUES !yes, again
SYSMAN> @USR$SCRATCH:[SYSMGR]AFILE.TMP
SYSMAN> EXIT
If the queue file went bad during running, then probably most queue
managers are still running. In that case, you were probably able
to recover the queue definitions using the procedure in
``Recovering old queue file'' section
.
So you now have a new queue manager running on one node,
and the old one still running on many other nodes, but talking
to a corrupt queue file. We need to stop all those old versions,
and restart them all to point to the new queue files.
Create a temporary file USR$SCRATCH:[SYSMGR]AFILE.TMP, and put the following lines in it:
$ STOP/QUEUE/MANAGERNow, run this file on all nodes in the cluster.$ @CMN$STARTUP:STARTQUEUES
$ SYSMANSYSMAN> SET ENV/CLUSTER
SYSMAN> SET TIMEOUT 0:5
SYSMAN> DO @USR$SCRATCH:[SYSMGR]AFILE.TMP
SYSMAN> EXIT
Keywords: RDCS, WARNER, DISNEY, controls, queues
Distribution:
D Baddorf
J DeVoy
B Kramper
T Watts