HA-LVM vs CLVM

[Matt Iavarone]

On Tue, Feb 2, 2010 at 3:29 AM, urgrue <urgrue at bulbous.org> wrote:
> It's a bit overkill in the sense that, as far as I have understood,
> HA-LVM and CLVM imply the use of red hat cluster and its associated
> components, luci, ricci, rgmanager, dlm, quorum and maybe fence device
> configuration, etc.
> All I really need is for the active node to mark the VG as "reserved"
> one way or another and have other node(s) respect that and thus not
> touch it unless forced to.
> I would assume (hope) there is some simple solution that I'm just not
> thinking of.
>
>
>
> On Mon, 01 Feb 2010 06:09 -0700, "Matt Iavarone"
> <matt.iavarone at gmail.com> wrote:
>> I haven't used ha-lvm, but it doesn't seem to be overkill for your
>> needs.  It looks tailored to your needs, in fact.  The kb says that
>> ha-lvm is for failover volumes, those that will only be mounted on one
>> host (http://kbase.redhat.com/faq/docs/DOC-3068).  So you don't need
>> to complicate things with a clustered file system.
>>
>> On Mon, Feb 1, 2010 at 1:36 AM, urgrue <urgrue at bulbous.org> wrote:
>> > Hello,
>> > I need to set up a simple failover scenario. The idea is to have two
>> > independent RHEL 5 systems on local disks, connected to a shared SAN.
>> > The application is all on the SAN. Only one node is ever running at a
>> > time. If the active node fails, the disk needs to be mounted on the
>> > passive node and the application started.
>> > Failover doesn't have to be fast or automatic, but it has to be simple
>> > and reliable.
>> > Depending on where I look, HA-LVM is sometimes recommended and other
>> > times it's CLVM. Looking at red hat cluster it seems HA-LVM is a bit
>> > overkill for my needs.
>> > Any suggestions or other options?
>> >
>> > --
>> > redhat-list mailing list
>> > unsubscribe mailto:redhat-list-request at redhat.com?subject=unsubscribe
>> > https://www.redhat.com/mailman/listinfo/redhat-list
>> >
>>
>> --
>> redhat-list mailing list
>> unsubscribe mailto:redhat-list-request at redhat.com?subject=unsubscribe
>> https://www.redhat.com/mailman/listinfo/redhat-list
>
> --
> redhat-list mailing list
> unsubscribe mailto:redhat-list-request at redhat.com?subject=unsubscribe
> https://www.redhat.com/mailman/listinfo/redhat-list
>

You don't need ricci or luci for either of these, but they do make it
easier to build and manage your clusters.  You can use
system-config-cluster in their place.

You can use, I assume, use just ha-lvm and gfs2 without rgmanager or
cman, but how will you manage the filesystem if the node fails?  Will
you manually mount it on your backup node?  And there are many options
for a fence device.  A red hat cluster using clvm and gfs2 is simple
and easy to manage.