Hi

I am learning (atleast trying to learn) HACMP on AIX 5.3. My current
setup is, 2 LPARs on the same physical frame, and I have HACMP
software installed.

My question is, do I have to have shared storage, and non-ip based
heartbeat?

This is not on a production box, so I don't really need true HACMP,
just a practical working one, to learn more.

Any ideas would be helpful! Thanks!

TIA

On Mar 29, 3:53 am, "Mag" <magaw...@gmail.com> wrote:
> I am learning (atleast trying to learn) HACMP on AIX 5.3. My current
> setup is, 2 LPARs on the same physical frame, and I have HACMP
> software installed.
>
> My question is, do I have to have shared storage, and non-ip based
> heartbeat?

Some form of shared storage is necessary if you want to configure
shared applications that run on either node (which is usually the
purpose of a cluster). The easiest way would be to set up another
LPAR in the same frame as a Virtual I/O server and export an
unused internal disk to both LPARs acting as HACMP nodes
(careful, in this configuration only a physical disk can be exported,
not an LV). This same disk can be used to set up a disk heartbeat
network (although this is not necessary for a simple test
configuration).

The other possibility for shared storage is, of course, physically
attached shared disks (SCSI or FC).

Regards,
Dmitri

On Mar 29, 2:14 am, "Dmitri Pasyutin" <pasyu...@gmail.com> wrote:
> On Mar 29, 3:53 am, "Mag" <magaw...@gmail.com> wrote:
>
> > I am learning (atleast trying to learn) HACMP on AIX 5.3. My current
> > setup is, 2 LPARs on the same physical frame, and I have HACMP
> > software installed.
>
> > My question is, do I have to have shared storage, and non-ip based
> > heartbeat?
>
> Some form of shared storage is necessary if you want to configure
> shared applications that run on either node (which is usually the
> purpose of a cluster). The easiest way would be to set up another
> LPAR in the same frame as a Virtual I/O server and export an
> unused internal disk to both LPARs acting as HACMP nodes
> (careful, in this configuration only a physical disk can be exported,
> not an LV). This same disk can be used to set up a disk heartbeat
> network (although this is not necessary for a simple test
> configuration).
>
> The other possibility for shared storage is, of course, physically
> attached shared disks (SCSI or FC).
>
> Regards,
> Dmitri


Dmitri:

Thats a good idea. What if my primary purpose is to share a NFS mount
point? Lets say I have a /mount_point, and if node 1 goes down, it
mounts it on node2?

Also, so its possible to assign the same physical disk to 2 different
LPARs on the same frame? I have 2 extra internals disks...
I am thinking something like this.
On the VIOs:
cfgdev
mksp vioc_tempvg_1 hdiskX
mkbdsp -sp vioc_tempvg 1G vioc_1_tempvg -vadapter vhost0 (to create
the backing device and attach the virtual device)

On client:
cfgmgr
mkvg newhDiskX tempvg

But I am still not sure if we can assign the same device, such as
hdisk to both clients, to emulate a heartbeat. Any advice?

On Mar 29, 3:56 pm, "Mag" <magaw...@gmail.com> wrote:
> Thats a good idea. What if my primary purpose is to share a NFS mount
> point? Lets say I have a /mount_point, and if node 1 goes down, it
> mounts it on node2?

Then the only resource you really need is a service IP address, to
which
your clients will connect and access the mount point. You don't need
any cluster disk resources (unless you explicitly want to set up a
disk
heartbeat).

> Also, so its possible to assign the same physical disk to 2 different
> LPARs on the same frame? I have 2 extra internals disks...
> I am thinking something like this.
> On the VIOs:
> cfgdev
> mksp vioc_tempvg_1 hdiskX
> mkbdsp -sp vioc_tempvg 1G vioc_1_tempvg -vadapter vhost0 (to create
> the backing device and attach the virtual device)
>
> On client:
> cfgmgr
> mkvg newhDiskX tempvg
>
> But I am still not sure if we can assign the same device, such as
> hdisk to both clients, to emulate a heartbeat. Any advice?

Like I said, you can only export a *physical* disk to 2 LPARs, not a
logical
volume. With mksp+mkbdsp you are exporting a logical volume. To export
a physical disk you would do something like this:

mkvdev -vdev hdiskX -vadapter vhost0
mkvdev -f -vdev hdiskX -vadapter vhost1

where vhost0 and vhost1 are server SCSI adapters corresponding to
node1 and node2, respectively, and hdiskX is the physical disk you
are exporting. After you run cfgmgr on both nodes, each one will see
a new hdisk which is physically the same disk. If you are only setting
up a disk heartbeat network and not creating any shared filesystems,
you don't need to create a VG over it.

Regards,
Dmitri