[Linux-cluster] HA-LVM Configuration & Snapshots

[Manish Kathuria]

Thank you for the response. I need a couple of clarifications.

On Mon, Feb 13, 2012 at 9:35 PM, Brassow Jonathan <jbrassow at redhat.com> wrote:
>
> On Feb 10, 2012, at 11:46 PM, Manish Kathuria wrote:
>
>> On a two node (active-passive) cluster with external storage on RHEL
>> 6.1, we need to use Snapshots for taking backup of the logical volumes
>> on the storage. As per the documentation, snapshots are not supported
>> on clustered logical volumes (using CLVM).  Since each of the logical
>> volumes on the external storage are going to be exclusively mounted
>> (and used) by a single node at any point of time, we can therefore use
>> HA-LVM instead of CLVM. However, for configuring HA-LVM also, there
>> are two suggested methods
>> (https://access.redhat.com/kb/docs/DOC-3068):
>>
>> - first one is the simpler one, using the CLVM variant while the other
>> one is the original method using volume_list and VG tags.
>>
>> I have a few queries related to both.
>>
>> Do snapshots work if we configure HA-LVM using CLVM  and run the clvmd daemon ?
>
> Yes (recently).  Whether you are using HA-LVM or CLVM, if the logical volume is exclusive, you can snapshot it.
>

As mentioned in the documentation, we have to make a logical volume
unavailable after creating using the "lvchange -an" command. Do we
need to use any similar commands or extra options  while creating the
snapshots ? The snapshots are going to be created and mounted on the
active node using scripts run through crond.

>> In case of the original HA-LVM configuration method, can we leave the
>> volume_list in  /etc/lvm/lvm.conf blank since we are not going to have
>> any shared VGs or LVs except for those which are going to be shared
>> using HA-LVM.
>
> You can't leave the volume_list blank.  Doing so would mean that your computer could not find your local drives the next time it boots up - forcing you to boot into an old kernel to bring the machine up.  You must at least put the local VGs in the volume list.

We are planning to use LVM only on the shared storage and not on the
internal devices on the nodes. The /boot, / and swap partitions would
be simple partitions of a hardware RAID device formatted with ext4 and
there would not be any local VGs.  That was the reason I asked this
question.

>  brassow
>

-- 
Manish