topic Re: Mount-f-noslot when mounting new shadowset in Operating System - OpenVMS

Mount-f-noslot when mounting new shadowset

Louis Nijman — Fri, 02 Mar 2007 11:10:00 GMT

Hallo,
I am new to this forum. I have the following problem:
When mounting a new shadowset on an Alpha ES45 using VMS V7.3-2 i get the error message:
%mount-f-noslot, no PCB available. Has anyone a suggestion as how to overcome this error? Any help will be appreciated.

Re: Mount-f-noslot when mounting new shadowset

Steven Schweda — Fri, 02 Mar 2007 11:25:38 GMT

alp $ help /message noslot

NOSLOT, no PCB available

Facility: SYSTEM, System Services

Explanation: The maximum number of processes that can be created at one
time has been exceeded.

User Action: Wait for another process to be deleted, or increase the
MAXPROCESSCNT system parameter.

How much don't you know about changing system
parameters like MAXPROCESSCNT?

Re: Mount-f-noslot when mounting new shadowset

Hoff — Fri, 02 Mar 2007 11:42:16 GMT

This means that you have too many processes running, or not enough process slots have been configured.

You've already been pointed to MAXPROCESSCNT, the other part of this is to look at what is present in the MODPARAMS.DAT file -- that's a common spot for tuning a system, and an equally common spot for mistuning a system -- and find out of it makes sense. I've seen all manner of parameter settings embedded there that are outdated (whether due to subsequent application or OpenVMS changes, or old information on system tuning), or should have been entered as MIN_, MAX_ or ADD_ parameter settings and that were erroneously inserted into MODPARAMS as absolute settings.

The contents of MODPARAMS can and do constrain what AUTOGEN can do to tune the settings, and what values it can establish.

In this case, I'd look at what's in MODPARAMS with a particular eye toward any process-level settings and any (absolute) setting of MAXPROCESSCNT, clean out any cruft, and where the values are still required I'd switch to MIN_, MAX_, or ADD_ where appropriate, and would AUTOGEN with FEEDBACK.

You can then begin an investigation of physical memory and process quotas. But the immediate NOSLOT limit can be lifted by using a MIN_MAXPROCESSCNT setting and an AUTOGEN (with FEEDBACK, and reboot) pass, or by reducing the numbers of currently active processes.

There is some risk to AUTOGEN, in that there are also folks that make system parameter settings changes directly, using SYSGEN or SYSMAN, and these changes can be lost over any subsequent AUTOGEN pass, and this AUTOGEN that can be required by a subsequent product installation, ECO, or OpenVMS upgrade. (Tweaking a parameter directly without also tweaking it in MODPARAMS is a common mistake. Most everybody makes it once.)

Ok, enough verbiage. Tweak MAXPROCESSCNT up via MODPARAMS and AUTOGEN FEEDBACK and reboot, or remove a few non-system, non-critical processes and continue running until you can schedule the tweak and the reboot.

Stephen Hoffman
www.HoffmanLabs.com

Re: Mount-f-noslot when mounting new shadowset

Jan van den Ende — Fri, 02 Mar 2007 13:21:30 GMT

First,

(now online) Louis: WELCOME to the VMS Forum.

Ok.

Background: I am a collegue of Louis, but because I will be incommunicado for nearly two weeks beginning next monday, we decided that it would be better to have someone else to deal with responses and further questions.

Now the problem.

- cluster environment, currently two nodes.
- current process counts just under & just over 1 K
- MAXPROCESSCOUNT both 3 K
- BALSETCOUNT both 2998
- 20 two-member shadow sets mounted clusterwide (20th set successfully mounted just minutes before)
- MOUNT is OK
- MOUNT/SYSTEM is OK
- MOUNT set /SHADOW= gives the error again (private MOUNT or /SYSTEM makes no difference)

The error is
%MOUNT-F-NOSLOT,
but HELP/MESS NOSLOT does _NOT_ mention MOUNT as a facility that should display the error.

Now, I at least was out of options.

Hope you can help.

Proost.

Have one on me.

jpe

Re: Mount-f-noslot when mounting new shadowset

EdgarZamora — Fri, 02 Mar 2007 14:17:51 GMT

It seemed strange to me that mounting a shadowset would cost you a process slot.

Have you checked your shadowing license? Maybe you don't have the traditional unlimited license and have a restriction as to number of shadow disks?

Re: Mount-f-noslot when mounting new shadowset

Hoff — Fri, 02 Mar 2007 14:49:00 GMT

>>It seemed strange to me that mounting a shadowset would cost you a process slot.<<

Gotta put that shadow server process somewhere.

Here's an old description of the overhead of shadowing...

http://h71000.www7.hp.com/wizard/wiz_4407.html

Re: Mount-f-noslot when mounting new shadowset

EdgarZamora — Fri, 02 Mar 2007 14:58:54 GMT

Yeah but the shadow_server was already out there. And it's only one process for any number of shadow sets.

I really think it's a license problem.

Re: Mount-f-noslot when mounting new shadowset

Jur van der Burg — Fri, 02 Mar 2007 15:02:27 GMT

You have more than 12 write bitmaps in the system, that's causing the ss$_noslot error.

Jur.

Re: Mount-f-noslot when mounting new shadowset

EdgarZamora — Fri, 02 Mar 2007 15:17:39 GMT

Neat. I'll have to keep that in mind as I'm just about to start setting up a large number of shadowsets on a new system.

Re: Mount-f-noslot when mounting new shadowset

Jan van den Ende — Fri, 02 Mar 2007 15:26:16 GMT

Jur,

do you mean that is a hard limit??

Sounds rather, or more than, extreme, seeing that write bitmaps are a rather recent devellopment, and the total of allowed shadowsets has for a very long time now been "limited" to 600 members, ie, 200 3-member sets or 300 2-member sets.

If the limit can be influenced, the gold medal question of course is:
The magic incantation, please, master?

But, at least it is the first explanation that is not at odds with the observations!

Proost.

Have one on me.

jpe

Re: Mount-f-noslot when mounting new shadowset

Robert Brooks_1 — Fri, 02 Mar 2007 20:02:33 GMT

You have more than 12 write bitmaps in the system, that's causing the ss$_noslot error.

--

There's a limit of 12 bitmaps per shadow set
(6 minicopy and 6 host-based minimerge),
but there is no limit systemwide (other than memory constraints).

I'll make sure the MOUNT maintainer hears about this one -- I'm curious to see who is actually returning NOSLOT.

-- Rob (ex-VMS Engineering, but still with HP)

Re: Mount-f-noslot when mounting new shadowset

Volker Halle — Sat, 03 Mar 2007 08:43:36 GMT

Louis,

did you check for existing device bitmaps on the shadowset(s) ?

$ SHOW DEV/BITMAP D

If there are old device bitmaps hanging around, you can delete them with $ DELETE/BITMAP xxx

The WBM facility (Write Bit Map) also reports errors into OPERATOR.LOG, did you check there ?

SS$_NOSLOT is being return with the WBM code (no more WBME entry slots available). This information should be added to the NOSLOT message help text.

Volker.

Re: Mount-f-noslot when mounting new shadowset

Jan van den Ende — Sat, 03 Mar 2007 10:36:23 GMT

Volker,

>>>
did you check for existing device bitmaps on the shadowset(s) ?
<<<

The drive was INITed just seconds before the first attempt to mount it, to create an entirely new, previously non-existing shadow set.

>>>
The WBM facility (Write Bit Map) also reports errors into OPERATOR.LOG, did you check there ?
<<<

We did not do that, Will do monday, and add that info.

>>>
SS$_NOSLOT is being return with the WBM code (no more WBME entry slots available). This information should be added to the NOSLOT message help text.
<<<

Not entirely clear what exactly you mean here (probably my lack of knowledge)
The reported error was MOUNT-F-NOSLOT. No secondairy message, nothing about WBM or WBME.

If Jur and Robert are indeed on the right track, then I agree, an addition to the HELP/MESSAGE text would be appropriate.

Proost.

Have one on me.

jpe

Re: Mount-f-noslot when mounting new shadowset

Volker Halle — Sat, 03 Mar 2007 10:56:54 GMT

Jan,

MOUNT itself does not return SS$_NOSLOT. The WBM (Write Bit Map) routines in the IO subsystem may return this error. The WBM-routines are called when mounting a shadowed disk. They also issue a message via OPCOM, as you will most probably find out on monday.

Volker.

Re: Mount-f-noslot when mounting new shadowset

Jan van den Ende — Mon, 05 Mar 2007 07:25:09 GMT

Okay, new (negative) results.

$ DIR/DAT OPC$LOGFILE_NAME
:OPERATOR.LOG 20-FEB-2007
Other node similar.
$ SEARCH those for DSA23 (intended shadowset) NOMATCHES
SEARCH for noslot -- no matches
search for mount-f- -- no matches

Any more ideas?

Proost.

Have one on me.

jpe

Re: Mount-f-noslot when mounting new shadowset

Volker Halle — Mon, 05 Mar 2007 07:39:41 GMT

Jan,

try searching for 'WBM'

Volker.

Re: Mount-f-noslot when mounting new shadowset

Jan van den Ende — Mon, 05 Mar 2007 07:46:18 GMT

Volker,

$ search
SEARCH-I-NOMATCHES

Proost.

Have one on me.

jpe

Re: Mount-f-noslot when mounting new shadowset

Jan van den Ende — Mon, 05 Mar 2007 08:05:49 GMT

Oops.

last reply should of course have been

$ search wbm

Proost.

Have one on me.

jpe

Re: Mount-f-noslot when mounting new shadowset

Volker Halle — Mon, 05 Mar 2007 09:15:17 GMT

Jan,

I have seen WBM OPCOM messages like this:

Message from user WBM$SERVER
%WBM-I-WBMINFO No more bitmaps slots available

Does the problem still exist ? If so, try REPLY/ENABLE and repeat the failing mount command.

Volker.

Re: Mount-f-noslot when mounting new shadowset

Jan van den Ende — Mon, 05 Mar 2007 10:32:24 GMT

AIAIAIAIAIAI !

I _HOPE_ we have found the cause.

It seems that the system was configured by an overly "economic" person, who wanted to save a few KB of memory (on 4 GB systems!)

SYSGEN param SHADOW_MAX_UNIT is set as 20 !!!

It looks like a reboot cycle coming up.

As an aside:

I really would have appreciated if the error message had been more relevant, ie, more in line what we have come to expect from VMS!

After the reboots we will report the results.

Proost.

Have one on me.

jpe