topic Re: Sysdump.dmp in Operating System - OpenVMS

Sysdump.dmp

UareTERMINATED — Tue, 04 May 2004 09:48:40 GMT

Hello Everyone!

IT is my first message in that forum. I wuld like t thank you first because it is a usefull forum. VMS is hard to manage.. Anyway let's talk about my problem

I have a 9node HPS system made with ES40 alphaserver.
The OS on It it OpenVMS 7.2-1 (i know it is old but i have no choice)

We had some problem on one of the node of the cluster and we decide to reboot it.

While the node boot we saw the following message
sys$system:sys$dump.dmp status= 910 (cannot find file)

But the file is there but the date is old
dir SYSDUMP.DMP;1 /date /size

Directory SYS$SYSROOT:[SYSEXE]

SYSDUMP.DMP;1 490518 29-NOV-2002 14:42:31.86

Total of 1 file, 490518 blocks.

My questions are :

1/ how this file is create? IS there any process who manage it(like for exemple the ERRFMT process manage the errorlog.sys)?

2/ is the date okay? because we are now in may 2004 and the system rebooted

3/How can we exalin the error during the reboot (my friend google didn't help ;0) and also all documentation on the web site)

Thank you very much for your help

Re: Sysdump.dmp

Lokesh_2 — Tue, 04 May 2004 10:35:29 GMT

Hi,

Check out Managing page,swap and dumpfiles in the following documentation...

http://h71000.www7.hp.com/doc/732FINAL/aa-pv5nh-tk/aa-pv5nh-tk.HTMl

HTH,
Thanks & regards,
Lokesh

Re: Sysdump.dmp

Martin P.J. Zinser — Tue, 04 May 2004 10:43:09 GMT

Hello,

Welcome here ;-)

1.) No special process. If you want to examine the dump file SDA is your friend.

2.) The Date is ok, you create the dump file once, either using sysgen or autogen.

3.) Check the settings of DUMPFILE_DEVICE in modparams.dat, the sysgen parameter dumpstyle and the output of f$getenv("dump_dev") and please report them back here.

Greetings, Martin

Re: Sysdump.dmp

UareTERMINATED — Tue, 04 May 2004 10:48:25 GMT

oki doki!! Great thanks for your fast reply..

I'll have a look and keep you update tomorrow

Re: Sysdump.dmp

John Eerenberg — Tue, 04 May 2004 10:57:15 GMT

> "Hello Everyone!"
Hi.

> "IT is my first message in that forum. I wuld like t thank you first because it is a usefull forum."
You are welcome. Please post more.

> "While the node boot we saw the following message"
> "sys$system:sys$dump.dmp status= 910 (cannot find file)"
>
> "But the file is there but the date is old"
> "dir SYSDUMP.DMP;1 /date /size"
>
> Directory SYS$SYSROOT:[SYSEXE]
>
> SYSDUMP.DMP;1 490518 29-NOV-2002 14:42:31.86
>
> Total of 1 file, 490518 blocks.

> "1/ how this file is create? "
When you install the system, autogen creates this dump file. But it almost always sizes it wrong. You only need to size sysdump.dmp once unless you add memory (sizing is a short topic unto itself.) Subsequent use of autogen, may alter the size further so I put a line such as "dumpfile=0" in modparams.dat to avoid this behavior.

> "1/ IS there any process who manage it(like for exemple the ERRFMT process manage the errorlog.sys)?"
As mentioned above, you can run autogen and unexpectedly change the size and create multiple versions of sysdump.dmp.
The answer to this question is that there is no detached process like errfmt. You are the one that manages sysdump.dmp.

> "2/ is the date okay? because we are now in may 2004 and the system rebooted"
It is fine. The size of the file is the important thing and where the dump file lives is too (yet, another topic).

> "3/How can we exalin the error during the reboot (my friend google didn't help ;0) and also all documentation on the web site)"
> Directory SYS$SYSROOT:[SYSEXE]
>
> SYSDUMP.DMP;1 490518 29-NOV-2002 14:42:31.86
Your post indicates a 9 node cluster. Your boot time error message indicates that the dumpfile is missing for one of the nodes.
From your directory display, notice that sysdump.dmp is in sys$sysroot. This is a unqiue directory for each node in the cluster.
Suggestion: You'll need to do the directory command from the same node that is giving the error message.
Better yet, you can display *all* the sysdump.dmp files by using the following directory commnand:
$ dire/date/size=all sys$sysdevice:[sys*.sysexe]sysdump.dmp

This should be enough information to solve the problem. If it isnt' . . .

To be absolutely complete, you'll need to get what are called dump_dev, bootdef_dev, and boot_osflags from each Alpha's console (the console being the >>> prompt when you shutdown/powerup a system). They are referred to as environment variables; these are just three environment variables and many more exist. In addition, you'll need to get the sysgen parameter dumpstyle from each node. Do the following for each node in the cluster preferably on OPA0 using the system account:

You can retrieve dump_style, bootdef_def, and boot_osflags by using the following commands (it should work on V7.2-1).
$ a=f$getenv("dump_dev")
$ sho sym a
$ a=f$getenv("bootdef_dev")
$ sho sym a
$ a=f$getenv("boot_osflags")
$ sho sym a
You can get the dumpstyle parameter by using the following commands:
$ mcr sysgen
use current
sho dumpstyle
exit

Re: Sysdump.dmp

John Eerenberg — Tue, 04 May 2004 11:09:55 GMT

Don't you love it when three people post at the same time???
But seriously, if you have several nodes in the cluster, this could be a simple case of a node not being setup.

Re: Sysdump.dmp

Wim Van den Wyngaert — Wed, 05 May 2004 04:05:02 GMT

During the startup of a node, sys$startup:clue$startup.com is executed.

It will analyze the dump file and put the result in sys$common:[syserr]. (Unless you defined clue$collect manually).

It also keeps a history of all crashes !

Re: Sysdump.dmp

UareTERMINATED — Mon, 10 May 2004 08:47:39 GMT

HI Again

Sorry for the delay but i was busy by other incident. ;0)

anyway, i read carefully your replies guys . Thanks

But i checked on all 9 nodes and the file is present on the 3 Back end (in cluster) and 4 front end (sorry it is a 7 nodes instead of 9 node, no billing node anymore)

But i saw a reply wich said : May be it is becausete file is not present on one of memeber of my 7 node..

IS it because, even i'm log as system account, i can't do from Backend node the followng command

dir "Front end name"::sys$sysdevice:[sys*.sysexe]sysdump.dmp

i got the follwing message
SYSDUMP.DMP;1 insufficient privilege or object protection violation ?????

Re: Sysdump.dmp

Uwe Zessin — Mon, 10 May 2004 10:55:47 GMT

Well, I guess you did not specify a privileged username/password, so the process on the remote site is very likely FAL$SERVER or DECNET. Those are non-privileged users and SYSDUMP.DMP is usually protected against users looking for useful information.

Re: Sysdump.dmp

Uwe Zessin — Mon, 10 May 2004 11:06:12 GMT

On second thought, it could also be that the protection of the master file directory (MFD= [000000]000000.DIR;1) is W:E. In that case your non-privileged network process cannot do a wildcard lookup of SYS$SYSDEVICE:[000000]SYS*.DIR. It would be nice to see the complete error message.

Re: Sysdump.dmp

Jan van den Ende — Mon, 10 May 2004 13:46:42 GMT

Uwe:

I think that this _IS_ the whole error message: the rest of the info about what is not allowed to be seen is not displayed BECAUSE it is not allowed to be seen!

UareTerminated:

You are accessing your 'remote' system with the default network access.
If you have set up proxy access, then you are using the defined proxy user (usually the same user as on the accessing system), but if not, then at the remote node you are the default network user (and obviously that is NOT blocked at your site!).
If you DO want the info to be supplied by the remote node, then use
$

OTOH, I understand this IS a cluster?
Then the info is available locally, and you can simply
$ dir sys$sysdevice:[sys*.sysexe]sysdump.dmp
If you heva multiple system disks (which I don't suspect) then replace 'sys$sysdevice' with the name(s) of the system disk(s) as seen from your local system.

Alternatively,

$ MCR SYSMAN
set environment/cluster
do dir sys$sysdevice:[sys*.sysexe]sysdump.dmp

But then, for every system booting off the same disk, you get the info over again.

If none of this helps, you make me very curious of your configuration!

Success,

Jan

Re: Sysdump.dmp

Mobeen_1 — Mon, 10 May 2004 23:46:31 GMT

Well, i don't think that OpenVMS is all that difficult to manage compared to some other OS. Its relatively very easy to manage VMS systems :-). I know that some times it tends to get difficult finding help if you are looking for anything. But thats why we are all here, to help out each other.

I have couple of suggestions for you. I hope they help you in managing your systems better and also finding resources that are needed to do so

1. Make it a habit to look for any info
at the DSNLink (at the ITS tool). If you
don't have DSNLink on your systems, you
can always look for "How to ...." at
http://askq.compaq.com. I find this very
usefull.

2. Since you mention that you have a 7 node
cluster, i would suggest that you put up
a seperate disk as a dump device. It will
certainly help you in some cases.

3. Many of our friends have already
mentioned on how to set up the dumps.

Let us know if you have any specific questions and we would be glad to clarify

regards
Mobeen