adding newly acquired disk drives to PVG's in /etc/lvmpvg and max size of file system

[Chris Gromala]

I've just purchased an additional disk enclosure and 4 disk drives for an N class server. I would like to extend both the LV and one of existing files systems to these newly acquired drives. The LV that I wish to extend to these drives was created using the Distributed Allocation Policy with extent based mirrored stripes. Do I simply add the new drive addresses to the PVG's in the /etc/lvmpvg and then perform the necessary commands? And if that is the case what would be the necessary commands?
Secondly, what is the max size of a file system these days? The current file system that I wish to extend is 41GB, I would like to increase the size to nearly 78GB. Is that possible? Any help would be greatly appreciated.

[James R. Ferguson]

Hi Chris:

First, filesystems under 11.x JFS can be as large as 1TB.

As far as extending your stripping to your new volumes, you can't do this. You will need to offload the data, create the volume group from scratch, and reload the data. See this thread for some comments.

http://forums.itrc.hp.com/cm/QuestionAnswer/1,1150,0x656e37f45ef7d4118fef0090279cd0f9,00.html

...JRF...

[Daniel Proulx]

I tend to disagree with James on 2 points:

1. Need to create "volume group" from scratch.

If something has to be recreated from scratch it would be the logical volume after updating the physical volume
groups.

2. Extending your stripping to your new disks cannot be done.

Extent based stripping (lvcreate -D y) is different from what I would call standard stripping (lvcreate -i nb_disks). I know you cannot extend an LV from new disks if this LV is using standard stripping.
Here is what I would try to extend your existing volume without destroying it first (I don't have a machine available with both free disks and LVM mirroring in order to test what I suggest):

- I assume you already have 2 PVGs, let's call them pvg1 and pvg2, and that your LV has the allocation
policy set to "PVG-strict/distributed" (lvcreate -s g -D y).
- Create 2 new PVGs, pvg3 and pvg4, each containing 2 new disks.
- Extend your LV from these new PVGs:

lvextend -L new_size lv_path pvg3 pvg4

I would not recommend adding the new disks to pvg1 and pvg2, because lvextend will try to distribute the
new required size from all the PVs in a PVG no matter how much space is currently allocated from them. If
you don't have space left on the existing disks you will not be able to extend your LV. If you have space left
on your existing disks you could extend the LV by the size equal to ( N * smallest free size on a disk in the
PVGs) where N is the new number of drives in one PVGs.

What do you think James?

Daniel.

[James R. Ferguson]

Hi Daniel (and Chris):

Daniel, I believe you see what I didn't.

1: You're correct that I said "volume group" when I should have said "logical volume". I was thinking simplisticly of one LV on one VG.

2. I hadn't thought about defining a second physcial volume group (PVG); rather I had imagined adding the new disks to the existing ones. In that case, my thinking was that the lvextend would fail for the very reason you cited.

I don't have a system on which to try this either.

Thanks for offering your thoughts! I think, in retrospect, that you are correct.

Regards!

...JRF...