make_recovery failure

[Dave La Mar]

Weekly we perform -
make_recovery -C -A -v -d /dev/rmt1mn
This week we are getting 33 of the following -
note: vgdisplay -v unsuccessful
This is followed by a core dump.
Has anyone seen this?
A little insight would be appreciated.
Support has not helped so far.
Thanks for any input.
dl

[Alan Wyskowski]

Try doing a vgdisplay -v on some of your volume groups and see if you get any errors and what they are for starters.
vgdisplay -v /dev/vgsomevolumegroup

[Craig Rants]

Dave,
What does a vgdisplay -v vg00 do?

If that fails it is not an ignite problem. Additionally, are there any other failures in your ignite log?

C

[James R. Ferguson]

Hi Dave:

What's changed this week?

Try, as root, at command line, doing"

# vgdisplay -v

...does that fault, too?

Regards!

...JRF...

[Dave La Mar]

I can display all vg's that are in /dev
except for three that are not active on this machine. (MC S/G clusters)
????

[Dave La Mar]

JF-
There was some lvm work done but the volume groups involved were exported and any entries in fstab commented out.
dl

[Craig Rants]

Dave,
What o/s are you using, anything other than 11.11 should only be concerned with vg00 only. So your serviceguard stuff really should not matter.

[Dave La Mar]

CR -
We're running 11.0. The three volume groups not active (MC/SG) have been there through other weekly successful make_recovery jobs.
dl

[harry d brown jr]

Your vg's were exported but not removed. They are still in lvmtab. do a man on vgscan, and follow the instructions for receating lvmtab:

mv /etc/lvmtab /etc/orig_lvmtab
vgscan

live free or die
harry

[Craig Rants]

Dave,
Look at this thread, I guess you were not the first person to encounter this problem.

http://groups.google.com/groups?hl=en&threadm=39cf4e04%240%244882%40businessnews.de.uu.net&rnum=1&prev=/groups%3Fq%3Dvgdisplay%2Bunsuccessful%26hl%3Den%26rnum%3D1%26selm%3D39cf4e04%25240%25244882%2540businessnews.de.uu.net

Good Luck,
C

[Dave La Mar]

HB -
This worked. Unfortunately, the new lvmtab does not caontain all the alternate link information contained in the original.
Support has noted we can keep this tape for disaster, but need to put back the original lvmtab file and work out the problem.
Any other suggestions?
dl

[Alan Riggs]

Sure -- just keep the new lvmtab and add back in alternate links (using vgextend) as desired.

[James R. Ferguson]

Hi Dave:

If you have executed a 'vgscan', READ the man pages (1M) for it. In addition to adding and/or adjusting alternate links, you need to make sure that your boot information is correct.

Regards!

...JRF...

[Dave La Mar]

Here's the result people -
1. We were misinformed. The new lvmtab does contain the alternate links.
2. The new lvmtab is missing only the vg that is part of a S/G cluster. This is the fail to machine.
3. A vgimport of the missing vg puts the lvmtab file back to the original size and entries.
4. A repeat of make_recovery once again fails.

The link to google may be correct with the reference to large files.
Will run this by support.
Still open to suggestions.
Thanks.
dl

[James R. Ferguson]

Hi Dave:

Since you indicate that 'vgdisplay' operates successfully at commandline, but appraently not with an Ignite 'make_recovery' try using a current Ignite version and 'make_tape_recovery' (the replacement for 'make_recovery'.

Download a current set of Ignite software from the Ignite site. No reboot is required for (re)installation:

http://www.software.hp.com/products/IUX/

Use this, in lieu of the old 'make_recovery', to capture a recovery image of all of vg00. The '-I' option has the *same* meaning as the '-i' option of the old 'make_recovery' -- it cause the Ignite process to be interactive when booting from tape:

# make_tape_recovery -x inc_entire=vg00 -I -v -a /dev/rmt/0mn

Regards!

...JRF...

[Dave La Mar]

JF -
In lieu of a solution, Support is suggesting the same thing.
It will be some time before we try this as we are in the midst of a critical time in our conversions.
We have narrowed the problem to the 600+ gb db that resides on another machine and is only in the lvmtab due to MC/SG support.
Our workaround will be to create the make_recovery minus the large vg in lvmtab then
mv the lvmtab containing it back into place following a successful make_recovery.
If anyone can see another solution, we would be much appreciative.
dl

[James R. Ferguson]

Hi Dave:

Remember that downloading and installing a new version of Ignite is quite fast. The 'swinstall' takes just a minute or so, and no reboot is necessary. You can then test the 'make_tape_recovery' directly. Please keep us informed. Thanks!

...JRF...

[Dave La Mar]

JF -
The latest version of ignite was installed. Using the syntax you provided make_tape_recovery appeared to work successfully though we still get vgdisplay -v errors.
The -C option is not available using the make_tape_recovery. This allowed a check point for check_recovery and I used the log file to notify the adms of success or failure via email.
Being new to hp-ux I'm not sure of all the options we should be using, but note:
we were running -
/opt/ignite/bin/make_recovery -A -C -v -d /dev/rmt/1mn
The man pages indicate the -A should not be run if one has very large vg's which we do. I tried running make_recovery without -A and it still failed.
Since we backup home directories and all db's reside on external storage with separate backups it looks like the syntax you provided for make_tape_recovery is adequate for our needs.
Thanks to all for the input.
dl

[Patrick Wallek]

The typical make_tape_recovery command that I use is:

/opt/ignite/bin/make_tape_recovery -a /dev/rmt/0mn -I -v -x inc_entire=vg00

Where the options are:

-a /dev/rmt/?mn - No-rewind tape device to use
-I (not this is an UPPER CASE I, not a lower case L) - When booting from the tape, boot into the Interactive installation menu
-v - Verbose output
-x inc_entire=vg00 - Include the ENTIRE vg00 volume group.

You are correct in that the -C option is not available with make_tape_recovery. That option will be going away eventually, at least that's the last I heard.

[Dave La Mar]

Very good. Thanks Pat for the confirmation.
To all that responded -
I have updated my email scripts to tail the new log - /var/opt/ignite/recovery/latest/recovery.log
and thanks again for the input.
dl