Shudown hangng on VAX 4705 at Que Manager Shutdown

[Warren G Landrum]

Guys,

When trying to shut down a VAX 4705 running VMS 6.1 (Iknow, I know) at a client site, system shutdown hangs at the point where it stops the que manager. Sending CTRL-Y interrupt makes the shutdown process continue.

Couple questions.

1. Anyone run into this and know of any fixes and/or workarounds?

2. Anyone know how to send a CTRL-Y so that if we had the shutdown going in a com file, I could just maybe wait for 10 or 15 seconds, and then send it to keep the shutdown going?

3. Any other suggestions?

Thanks,

Warren

[Hoff · Accepted solution]

1: Yes. The have been various hangs in this area eons ago. There can be problems with an ill-configured queue manager, or with an errant host specification, or with shadowing, or elsewhere. Load up the "current" ECOs, and see if it works.

Poke around within the shutdown (verify) and find out exactly what is hanging -- you might be able to work around this by short-circuiting whatever failed and locked up.

2: hit break on the console, and hard-halt the box. There's no way to send a ^Y from within DCL; that character has to arrive in-bound from a terminal or a terminal emulator or console management package.

3a: Consider fixing the "I know, I know" part you refer to, at least as far as applying the ECOs for V6.1 (and those are themselves rather old now) and preferably V6.2, or V7.3. That'll get the known problems resolved.

3b: Look at your data archives and ensure you can restore what you've archived, as a VAX 4000 model 705 series has increasingly old and increasingly failure-prone storage. I'd be looking at replacing the whole box; based on the release schedule, this configuration hasn't seen significant changes in a dozen years -- this software is older than Windows 95, after all. Hardware does eventually fail. SIMH or CHARON-VAX, or a move to an Integrity, are the obvious paths forward.

[Martin Hughes]

Are you able to stop the queue manager generally? I.E. is the problem only evident when shutting down the system?.

I take it this is a standalone node.

[Thomas Ritter]

Warren, does the system eventually shutdown or just hang indefinitely ? How long would it take to shutdown without the ctrl/y ?
As previously asked is the node a cluster member ? If so where was the queue manager running ?

[Wim Van den Wyngaert]

2. For this purpose I execute everything that is unreliable in a spawn with a timeout. The command will continue until the end of startup/shutdown and will be killed if not yet ready then.

$ @limit_spawn 30 stop/que/man

$! LIMIT_SPAWN
$! ===========
$! This procedure executes a command and waits a limited time for its
$! completion. The status of the command is NOT returned !
$!
$! P1 time to wait for completion, default 5 seconds
$! P2...P8 command to execute
$!
$! THIS FILE MAY NOT BE MODIFIED OUTSIDE THE SYSMGT MAINTENANCE SYSTEM
$!==============================================================================
$ say = "write sys$output"
$ p2=p2 + " " + p3 + " " + p4 + " "+ p5 + " " + p6 + " " + p7 + " " + p8
$ proccnt=f$getjpi(0,"jobprccnt")
$ spawn /nowait /nolog /input=nl: 'p2'
$ if p1 .eqs. "" then p1=5
$ cnt=0
$wait:
$ wait 00:00:01
$ cnt=cnt+1
$ if cnt .gt. p1
$ then
$ say "!!! ATTENTION : ''p2' didn't complete in ''p1' seconds !!!"
$ exit
$ endif
$ if proccnt .ne. f$getjpi(0,"jobprccnt") then goto wait
$ exit

Wim

[Warren G Landrum]

Thanks All.

Hoff, I mis-spoke. Architecture WAS 4705, but it is now running on a CHARON-VAX. We are actually in process of testing /validating before going live and that's where this problem is occurring.

Martin - Yes, can stop que manager interactively. Only been attempting to do so during this migration testing, tho.

Node is standalone on CHARON-VAX. WAS part of 2-node (4705s) VaxCluster.

Thomas - Hangs indefinitely w/o sending CTRL-Y

Wim - Will look at your spawned job, altho I guess the Real question I have is Can you automate a shutdown w/o using batch ques? I was trying to think through it, and I always come to the point that you can't do it in a batch que, since when SHUTDOWN stops the que manager, the batch job would abort and the shutdown process would be unfinished at that point. I've just never really tried to automate shutdown. Always did it interactively either executing the @sys4system:shutdown or by typing the SHUTDOWN symbol.

[Hoff]

Load the ECO kits for V6.1, and (if you can at all get there) the V6.2 release and its ECOs (that release is the next one forward that is in Prior Version Support status), or forward to the "current" V7.3 release.

There are no major releases from V6.0 forward on OpenVMS VAX, and both user-mode and kernel mode applications are expected to be fully forward-compatible.

If it still hangs during shutdown (on V6.2 with ECOs or V7.3 with ECOs), then the differential involves whether the hang happens with a real VAX 4000 model 705 box. If it does (and you have a support contract), the report goes to HP. If not, the report goes to SRI.

[Warren G Landrum]

Hoff,

Moving off of VMS 6.1 does not appear to be an option, but I WILL check to see if it has latest ECOS loaded and if not, I will recommend to them that we do so and see if this rectifies this particular problem.

Thx,

w

[Dean McGorrill]

hi Warren,
you can run a detached job set off by
a batch queue to shut your system down. I
look to see if I can find the stuff we
used to use tonight. Dean.

[Warren G Landrum]

That would be great Dean.

Thanks in advance !!!

w