reducing a filesystem to allocate to another file system

[James Ellis_1]

I need to (on prod box) reduce the /dbawrk filesystem in order to have more space and allocate this space to /oracle for the Oracle 9i upgrade. I am getting error msg like errno 16, and believe this is because we have OnlineJFS v3.1 and doing a fstyp -v /dev/vg04 /dev/dsk/dbawrk shows filesystem is version 3. Since I don't have version 4 (which indicates its a OnlineJFS v3.3), and this is a production box, I thought it may be better to just blow away the /dbawrk (currently at 13.4G), and recreating the /dbawrk with 10G. I will then extend /oracle with the 3.4G space. By the way /dbawrk and /oracle are from the same vg.

I have a couple questions, we can easily restore from a weekly backup, but is it tricky to recreate the mountpoint with a smaller size? Anything I should look out for?

Thanks.

[Robert-Jan Goossens · Accepted solution]

Hi,

You are correct, blow away the filesystem and recreate it with a smaller size.

# lvremove
# lvcreate
# newfs
restore data from backup.

Hope this helps,
Robert-Jan

[Kent Ostby]

You could also just:

unmount
lvreduce
newfs

[Sridhar Bhaskarla]

Hi James,

I would take the approach based on how much /dbawrk filesystem is used. If it is only used a little (say around a gig or so), then I will temporarily create a filesystem of that size, move the data, recreate the /dbark with less size and move the data back. If you don't have free space in some other volume group enough to take the backup of /dbawrk, then I will go with your approach.

I try to depend less on tapes and backup.

-Sri

[James Ellis_1]

All ideas are good.

Sridhar, The used file system is a tad over 2G. I don't have 2G laying around to create a temp filesystem. Even the /tmp is 1.1G available.

I had thought I could just backup the whole /dbawrk on tape, and then blow away the mountpoint and recreate the mountpoint and restore from that tape. But I had problems restoring from tape before, in that the directories were all wrong after the restore. I presume this is because I did not explictly say where to restore to?

Thanks.

[Jeroen Peereboom]

James,

using fbackup, tar or cpio you should not have problems restoring the files. The destination is the same. But since you have had problems before, make 2 backups and test them before removing the files.
I do not know about Online JFS, but I thought OJFS is meant for situations like this. Shouldn't it work, reducing a filesystem and lvol?

JP

[Sridhar Bhaskarla]

James,

2 GB is not big deal to put it on tape. But you still have options doing an 'scp -rp' or 'rcp -rp' or like mounting an NFS filesystem to a remote system where disk space is available etc., I am suprized to see you don't have a filesystem at /var/adm/crash with this being a production server.

In either the case, it is a good idea to prepare atleast two tapes if the data in this filesystem is too important. As I said, backup and restore from tape is my last option.

-Sri

[Sundar_7]

Hi James,

Since only 2 GB is used out of 13.4 GB, you could try to defragment the file system and try reducing it.

1) Take the backup
2) run fsadm -F vxfs -dD -eE /dbawrk
3) fsadm -F vxfs -b 10000m /dbawrk
4) lvreduce -L 10000m /dev/vg01/lvoldbawrk

See if this helps

Sundar