[Linux-cluster] Help needed for mysql cluster

[Fajar A. Nugraha]

On Thu, Aug 30, 2012 at 2:36 PM, Rajendra Roka <rajenddra at gmail.com> wrote:
> I don't mean to use mixed system. I am currently using 2 node redhat cluster
> in vmware with RDM. The reason behind this is high avaliabliy in MySQL. Not
> sure how Vmware able to handle the high availabily in this case?

Have you tried VMHA?

> How can I
> make 2 nodes MySQL servers in vmware  that runs in failover mode without
> using Redhat or DRBD cluster ??

It's one VM. vmware can take of moving the VM across any available
physical nodes in case one physical node goes down.

Similar in concept to RHCS, but instead of restarting/moving
application, you're restarting/moving VM.

>
> We want 2 nodes just in-case if we need to perform maintenance in one node
> another node takes over. We want identical copy in each box since we will
> not have direct access to iSCSI or storage.

Are you REALLY sure you don't have shared storage? Cause if so, the
only way to reliably have "identical" copy is:
- have DRBD emulate a shared storage, OR
- use mysql cluster, which is another beast altogether. MUCH more complex.

> The idea is to get rid the RDM
> device.

Why?

By removing RDM and shared storage, you're actually increasing the
level of complexity MUCH more.

> We don't want in load balance mode but master/slave works for our
> environment.

mysql has plain master/slave setup, without the need of any additional
clustering software, but there's no setup that I know of that can
RELIABLY promote a slave to master AUTOMATICALLY in case of failure.

Manual slave <-> master switch should work fine though, but it's
somewhat complicated.

>
> I found problem with DRBD implementation here since we have only one nic
> card(vlan) patched in this box.
>
> Please suggest.

Prioritize your options. Every change has it cost. You need to
determine whether the cost is worthed.

It seems to me you're very determined to change your current setup,
when in fact it works just fine and changing it will (most likely)
make it worse.

IMHO there's nothing wrong with NOT being able to vmotion a RHCS node.
You don't NEED to vmotion it. Just shut it down, perform whatever
maintenance you need on the physical server, and start it again. RHCS
should take care to move any managed resource to the surviving node.

-- 
Fajar