/var/adm/sw/save directory using up a lot of disk space

[kaushikbramesh]

Hi,

We are currently running HP-UX B 11.23 on a Itanium based rx7620 server. The /var partition has reached 100%. I see that the /var/adm/sw/save directory is taking up most of the disk space. This directory seems to store the information about all the patches installed on the system. However, we have another identical system with the same configuration (hardware and OS), I dont see this problem on this system. I'm not sure why this directory is using up so much disk space. And is the best way to clear this directory.

bdf - on system with var partition full
=======================================

dev/vg00/lvol3 1048576 235328 806936 23% /
/dev/vg00/lvol1 303125 107791 165021 40% /stand
/dev/vg00/lvol9 6291456 6291456 0 100% /var
/dev/vg00/lvol11 4194304 83514 3855885 2% /var/tmp
/dev/vg00/lvol10 8372224 18524 7831601 0% /var/adm/crash
/dev/vg00/lvol7 4407296 2676032 1717784 61% /usr
/dev/vg00/lvol8 524288 16597 475968 3% /usr/openv
/dev/vg00/ecc 524288 16632 503704 3% /usr/ecc
/dev/vg00/lvol6 1048576 430888 613824 41% /tmp
/dev/vg00/osmf 212992 7556 192654 4% /osmf/mgmt
/dev/vg00/xfer 1048576 94872 946440 9% /osmf/mgmt/xfer
/dev/vg00/sched 114688 5166 102732 5% /osmf/mgmt/scheduler
/dev/vg00/esm 524288 123881 375423 25% /osmf/mgmt/esm
/dev/vg00/eco 114688 1141 106458 1% /osmf/mgmt/eco
/dev/vg00/condir 114688 1141 106458 1% /osmf/mgmt/condir
/dev/vg00/UCPS 409600 97712 309640 24% /osmf/mgmt/UCPS
/dev/vg00/lvol5 4194304 2084624 2093216 50% /opt
/dev/vg00/lvol4 114688 9432 104472 8% /home

VVT-nuncas02-Oracle # du -ks ./*
0 ./cache.lck
8 ./cleanupfile.prev
8 ./defaults
8 ./defaults.patchfilters
8 ./getdate.templ
0 ./host_object
0 ./host_object_np
8 ./pre_update
82032 ./products
3632 ./queue
====================
1146792 ./save
====================
4176 ./save_custom
56 ./security
0 ./software
448 ./swagent.log
40 ./swagentd.log
360 ./swconfig.log
288 ./swinstall.log
8 ./swmodify.log
8 ./swremove.log
8 ./target_hosts
0 ./targets
8 ./tmp
0 ./ui
VVT-nuncas02-Oracle # pwd
/var/adm/sw

Thanks and Regards
Kaushik

[MarkSyder]

The command:

cleanup -c 1

will delete patches that have been superseded at least once (or you can err on the side of caution and replace the 1 with a 2).

It's worth looking in other places for large files. Usual candidates:

/var/adm/btmp
/var/adm/sulog
/var/adm/wtmp
/var/adm/cron/log
/var/adm/lp/log
/var/adm/syslog/mail.log
/var/adm/syslog/syslog.log

I would not recommend deleting these files as this might cause major problems: go into SAM and trim them to the recommended size. Any backup logs can be nulled (NOT deleted).

Mark Syder (like the drink but spelt different)

[Muthukumar_5]

Check the identical patch informations. If you are haivng there you can clear this. If you want then take a backup of this also.

You can use cleanup utility to clean patches also. see man cleanup

hth.

[LiPEnS]

Hi

You could run the cleanup command first in a preview mode:
# cleanup -p -c 1
and clean mode:
# cleanup -c 1
This clean superseded patches

Regards

[RAC_1]

That dir saves the installed pachets. In case you want to revert to old patch it can work. Generally cleanup -c1 will allow you to go back 1 level. cleanup -c1 -p ->to check what paches are superceded and can taken out.

I personally prefer havinf seperate FS for /var/adm/sw

[kaushikbramesh]

Hello,

Thanks for all your replies.
I dont seem to find the cleanup command on our boxes.

# which cleanup
no cleanup in /usr/sbin /usr/bin /usr/ccs/bin /usr/contrib/bin /usr/contrib/Q4/bin /opt/perl/bin /opt/ipf/bin /opt/hparray/bin /opt/nettladm/bin /opt/fcms/bin /opt/sec_mgmt/bastille/bin /opt/resmon/bin /opt/gnome/bin /usr/bin/X11 /usr/contrib/kwdb/bin /opt/mozilla /opt/wbem/bin /opt/wbem/sbin /opt/graphics/common/bin /opt/mx/bin /opt/sec_mgmt/spc/bin /opt/upgrade/bin /usr/contrib/bin/X11 /sbin /home/root /usr/opt/temip/bin /usr/opt/tns/bin /usr/opt/temip/man /usr/opt/tns/man /dboracle/orabase/product/9.2.0.2/bin

Regards
Kaushik

[Robert-Jan Goossens_1]

Hi Kaushik,

The cleanup command is delivered with patch PHCO_32220.

Patch Description: s700_800 11.23 HP-UX Patch Tools

Best regards,
Robert-Jan

[kaushikbramesh]

Hello,

I'm quite new to HP-UX. I'm puzzeled why this happened only on one machine. Is there anything I'm missing.

Thanks for all your suggestions.
Regards
Kaushik

[MarkSyder]

Several possible reasons why just this one box. First of all, I would suggest checking the log files I referred to earlier - if this is a heavily used box, they might be larger than on a less used box.

Mark

[kaushikbramesh]

Hello Mark,

As you suggested these log files seem to be quite large, I will trim these files and check again.

Thank you for your suggestion.
Kaushik

-rw------- 1 root other 187776 Nov 11 11:06 btmps
-rw------- 1 root root 144518 Nov 10 16:15 sulog
-rw-rw-r-- 1 adm adm 1736704 Nov 10 16:13 wtmps
-rw-r--r-- 1 root root 4710400 Nov 10 15:54 /var/adm/cron/log
-rw-r--r-- 1 root root 2357145 Oct 25 15:42 /var/adm/cron/OLDlog
-rw-r--r-- 1 root root 174448 Oct 25 15:42 /var/adm/syslog/OLDsyslog.log
-r--r--r-- 1 root root 104182 Nov 11 11:00 /var/adm/syslog/mail.log
-rw-r--r-- 1 root root 245760 Nov 10 14:38 /var/adm/syslog/syslog.log

[kaushikbramesh]

Hello,

I've managed to get the disk space back on the /var partition.
Yesterday we were running a few applications on this box, the applications had filled up some of the logfiles under the /var partition.
These processes were spawned off from swconfig.
However the session which was running the swconfig was killed for some reason and the processes were still in the ps table. The log files which had grown quite big in the var partition were deleted. These processes though had stopped logging to the logfiles in /var, bdf was returning /var partition as full. Now when I killed all these processes, bdf seem to show /var as 22%. I'm confused again :-( I dont understand why bdf was showing var as 100%.

Thanks in advance for all your suggestions.

Regards
Kaushik

[MarkSyder · Accepted solution]

If processes were running, the space from any files they had open would not be released until the processes were killed. This would apply even though you deleted the files.

Mark

[kaushikbramesh]

Thank you very much for all your suggestions and help.

Regards
Kaushik