Cleanup error

[George_Dodds]

Hi i've just ran cleanup to free up some space in /var on an old L class server we have running 11.00 and i got the following error

## Cleanup program started at 10/27/05 13:03:45
Commit patches superseded at least 1 time(s) on 'server1'.

ERROR: swlist(1M) encountered a problem displaying the installed software on 'server1'.
This problem must be corrected before continuing:

WARNING: Incomplete definition for the root (installation) "/",
beginning at line 0. The following attributes are not defined: products the products contained in the root
ERROR: The depot "server1:/" exists but is currently empty.

Is there anyway round this as i need to clear out the old patches to free up space.

Cheers

George

[Luk Vandenbussche]

Hi,

Have you another swinstall session running?
then close is first.

Try also

swconfig \*

[Muthukumar_5]

Try as,

# cleanup -c

see man cleanup

-Muthu

[George_Dodds]

No other session running and i'm getting the depot empty message for swconfig as well

[Luk Vandenbussche]

Restart your swagent daemon

/sbin/init.d/swagentd stop
/sbin/init.d/swagentd start

[Arunvijai_4]

# swagentd -r will help

-Arun

[Robert-Jan Goossens_1]

George,

Title: Incomplete definition for the root (installation)
Document ID: KBRC00003721

Europe
http://www4.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000080045540

US
http://www2.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000080045540

Regards,
Robert-Jan

[George_Dodds]

Still getting the same error

[George_Dodds]

On the right track Robert

but getting

ERROR: "lonpcs01:/": This operation will exceed the absolute limit
of disk space on 1 volumes.
* The analysis phase failed for "lonpcs01:/".
* Analysis had errors.

as /var at at 100% and i've cleared all of the logs down i can

[Andy Torres]

Here's a thread with a similar problem. The resolution was that someone either removed the complete /var/adm/sw/products directory or compressed files under the same. See if that helps.

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=702648

[George_Dodds]

Well the /var/adm/sw/products directory is there, i will check for any compressed files

[Andy Torres]

Looks like you're going fine now, but you may have to extend /var before continuing.

Try looking under /var/preserve or /var/tmp for big files to remove.

Run a du to list the largest files under /var and start going through the list:
du -akx /var | sort -nr | more

[George_Dodds]

Doh! my cough medecine is sending me daft.

The issue with the install was lack of disk space in /var.

should i compress some files in /var/adm/sw/products so i can run the install.

And once that is done then uncompress them again so i can run the cleanup script?

[Andy Torres]

No. I didn't express myself correctly earlier. You should NOT compress files under /var/adm/sw. That was their problem, not their resolution.

I'd look for large files and explore extending /var first. Try the du command above. The list can sometimes surprise you, and you may find other garbage files you didn't realize existed.

[George_Dodds]

I am unable to extend /var as this disk is full and i cannot get the downtime on the server.

We are going to migrate from this server onto a newer one with newer os and bigger disks soon anyway.

[George_Dodds]

I've cleared down all other files i safely can.

[Andy Torres]

Sounds like you may have dodged a bullet if you're migrating to another server soon. I hope it goes well.

If you find you have to load patches specific to the migration you may have to download just the few you need in order to get it done.

Good luck!

[George_Dodds]

The migration is a fresh build thankfully just transferring data, i just need to keep this server alive long enough till it is all completed. (could be months!)

but /var at 100% aint helping :)

[Charles McCary · Accepted solution]

George, your other option might be to build a directory on another filesystem that has some space, move the contents of the save directory to this new directory and then build a logical link from your new directory back to save, something like:

1) # mv /var/adm/sw/save /newfs/newdir/

2) # cd /var/adm/sw
# ln -s /newfs/newdir/save save

This should work, and you SHOULD be able to then run the cleanup program and move everything back, or just leave it as is until you move to new servers...

DISCLAIMER...I haven't tested this. :)

[Andy Torres]

If you've removed all files under /var that you can, your only option may be to plan some downtime. It may not seem like an option right now, but with /var at 100% it may be a matter of time. Better to plan your downtime than to explain an unplanned outage.

I'm not usually keen on "voodoo" solutions like moving files and linking, etc., in order to get past something like this. But maybe one of the other folks here can suggest something less severe.

[Charles McCary]

I really don't see the "voodoo" in the solution I offered, if no processes are hitting the save dir (and they shouldn't be), then there won't be an issue.

Worst case the link can be removed and the files moved back. No harm no foul.

I guess I don't see the mystery...

[George_Dodds]

Allthough Charles idea wasnt pretty or best practice it worked, i was able to install the void package and ran swlist ok.

Cleanup works fine now :)

Thansk for everyones input on this as it has given me a but more breathing space with the migration.

Cheers

George

[Andy Torres]

I wasn't criticizign your method, Charles. "Voodoo" is just my way of saying "workaround". As long as the impact is completely understood, I'd say have at it.

Sounds like it worked, so.... success! :-)

[Charles McCary]

COOL - glad to hear it.