Extending logical volume with new LUN's

[Brendan Newport]

I thought I knew LVM...ho hum!

I have a twin-node ServiceGuard cluster with two FC30 arrays seperated over a kilometre. Data on the primary (active) node's array is mirrored onto the other (redundant) nodes array.

I recently defined a new LUN on each array, and added the LUN's to a volume group (don't worry about the vg exporting and importing stuff, which has been done.)

I'd like to extend a few LV's in the same volume group, using the space from the new LUN, and ensure that new space is mirrored to the LUN in the redundant nodes array.

But I can't work out the syntax for the commands necessary.

Physical volume groups are established;

--- Physical volume groups ---
PVG Name nike_kl
PV Name /dev/dsk/c0t3d1
PV Name /dev/dsk/c0t3d3
PV Name /dev/dsk/c1t3d0
PV Name /dev/dsk/c1t3d2

PVG Name nike_ro
PV Name /dev/dsk/c1t4d1
PV Name /dev/dsk/c1t4d3
PV Name /dev/dsk/c0t4d0
PV Name /dev/dsk/c0t4d2

The logical volumes at present have just one LUN in use, i.e.;

--- Distribution of logical volume ---
PV Name LE on PV PE on PV
/dev/dsk/c1t3d0 11268 11268
/dev/dsk/c0t4d0 11268 11268 (this is the LUN on the array attached to the redundant node.)

The new LUN I want to use is defined as;

/dev/dsk/c1t3d2
/dev/dsk/c0t4d2 (the LUN on the array attached to the redundant node)

So the question is;

how do I extend a logical volume with the new LUN, and ensure that the new LUN is mirrored to the identical LUN on the redundant node's mirror, in other words, use /dev/dsk/c1t3d2, and ensure its extents are mirrored at /dev/dsk/c0t4d2, whilst retaining the current LUN already in use and its mirror?

In essence, I want to extend /dev/vgxx/lvolxx using /dev/dsk/c1t3d2 and ensuring extents on that LUN are mirrored to /dev/dsk/c0t4d2.

Thanks in anticipation,


Brendan

[James R. Ferguson]

Hi Brendan:

Try:

# lvextend -L /dev/vgxx/lvolxx /dev/dsk/c1t3d2

# lvexend -m 1 /dev/vgxx/lvolxx /dev/dsk/c0t4d2

...JRF...

[Laurent Paumier · Accepted solution]

If the lvol is already mirrored, you should be able to do it with :

lvextend -l /dev/vgxx/lvolxx /dev/dsk/c1t3d2 /dev/dsk/c0t4d2

[Carsten Krege]

You already setup physical volume groups. That is good, but you need to make sure that the logical volume uses the allocation policy pvg-strict (lvchange -s g /dev/vgXX/lvolY).

If you did this LVM will make sure that the master copy and the mirror copy are always located in different PVGs. Therefore it is enough to simply do the following:

# lvcreate /dev/vgXX
# lvchange -s g /dev/vgXX/lvolY
# lvextend -m 1 /dev/vgXX/lvolY
# lvextend -l /dev/vgXX/lvolY

If you want to be very sure that the extents are only allocated from the disks /dev/dsk/c1t3d2 and /dev/dsk/c0t4d2 you can specify them in the lvextend command:

# lvcreate /dev/vgXX
# lvchange -s g /dev/vgXX/lvolY
# lvextend -m 1 /dev/vgXX/lvolY
# lvextend -l /dev/vgXX/lvolY /dev/dsk/c1t3d2 /dev/dsk/c0t4d2

This will allocate the extents from these two disks only. You don't even need to set the PVG-strict allocation policy for this (strict(=default) is sufficient).

Carsten

[Brendan Newport]

Ah,
now all I needed was confirmation that my first thought was good. Laurent got there first, and lo, I have two logical volumes extended and allocation mirrored correctly.

Now of course, having done that, the DBA asks for a new logical volume to be created...

ho hum.

Thanks to all who contributed.