Reducing lvol in a MC/SG cluster

[Shokhat Ali Sheik_2]

What are the steps to safely reduce the size of a logical volume in an MC/SG cluster?

[Sanjiv Sharma_1]

Hi Ali,

Please check this:
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=109782

hth.

[Mark Grant]

You don't need to do anything particular because it is MS/SG unless you are removeing a disk. I'll assume you have onlineJFS so you can do this online.

Just make sure you have reduced the fileystem with "fsadm -b M /mountpoint" and then do a "lvreduce -L M /dev/vgXX/lvolX"

You don't need to re-export the map file to the other nodes.

If you are going to remove a disk from the volume group, you need to "vgexport -m mapfile -p -s vgXX" and then copy the mapfile to all the other nodes and do "vgimport -m mapfile -s vgXX"

[Shokhat Ali Sheik_2]

I did the following:
umount /util
fsadm -F vxfs -b 500 /util

I get the following message:
/util is not the root inode of a vxfs file sytem.
Why am I getting this.
There are only 1 logical volume in the vg.(/dev/vgutil/lvutil). This is being tested on a non SG system prior to doing the actual lvreduce in the cluster.
Current size is 1023.

[Karthik S S]

Unless otherwise you are not going to rename the LVOLs/VGs you need not worry about anything. I would advise to halt the package then manually activate the VG and do the necessary changes.

-Karthik S S

[Karthik S S]

Try,

fsadm -F vxfs -b 500 /dev/vgutil/lvutil

-Karthik S S

[Karthik S S]

I am sorry about that ...

If you planning to reduce the lvol then,

mount the filesystem
reduce the filesystem
reduce the lvol
remount filesystem

-Karthik S S

[Mark Grant]

fsadm works on mounted filesystems. If you don't have onlinejfs you will need to do more work than this. If you don't have OnlineJFS installed, I would backup, destroy the filesystem, reduce the logical volume and then restore.

If you do have OnlineJFS, then you need to have the filesystem mounted.

[G. Vrijhoeven]

Hi,

Fsadm is online so the filesystem must be mounted like others have mentioned. I like to add that it is good practice to make a full backup before reducing the filesystem. Most of the times it works fine, but not always!

HTH,

Gideon

[Shokhat Ali Sheik_2]

I Did the following on the test system:
fsadm -F vxfs -b 2000000 /util
umount /util
lvreduce -L 2000 /dev/vgutil/lvutil
fsck -F vxfs -o full /dev/vgutil/rlvutil

and mount -
All is OK - files are still accessable.

Thanks to all for your assistance. Now for the production system, tonight after backups!
If there are anything additional, will be much appreciated - thanks.

[Mark Grant · Accepted solution]

The only thing I would mention is that the -b option of "fsadm" is number of blocks. Your filesystem is unlikely to have 1024k size blocks. I would suggest you use the undocumented M for your fsadm to be sure your logical volume and filesystem are the same size.

fsadm -F vxfs -b 20480M /mountpoint

This wil give you a real 20GB filesystem size.