Mirroring, PV Links, and EMS HA Monitors

[Duncan Edmonstone]

I've been reading up on the disk monitors in the EMS HA Monitors product with a view to utilising this with MC/ServiceGuard. However there's one point that confuses me... the rules for using the pv_summary seem to contradict each other when using PV Links *AND* mirroring!

Imagine that I have a node in a cluster - the node has two HBAs connected via a SAN cloud (details unimportant) to two physical disks. The host might see the disks something like this:

HBA 1 0/1
|- c1t0d0 (first disk)
|- c1t1d0 (second disk)


HBA 1 0/2
|- c2t0d0 (first disk)
|- c2t1d0 (second disk)

Now lets say I want to create a mirrored logical volume on these two disks and have the volume group its in monitored by the EMS HA monitors. I look at the 'rules' I must follow, and first notice that:

"Mirroring must be PVG-strict.
Mirrored volumes must reside on a different bus from the original volume to avoid a single point of failure and to obtain the best pv_summary value for that mirror. This is done automatically by LVM if you created the PVGs while setting up mirroring."

OK, so that means I need an /etc/lvmpvg file that looks something like this:

VG /dev/vgtest
PVG mirrorA
/dev/dsk/c1t0d0
/dev/dsk/c2t0d0
PVG mirrorB
/dev/dsk/c1t1d0
/dev/dsk/c2t1d0

OK... but there are also some rules for using PV links:

"To use the HA Disk Monitor with ServiceGuard, PV links must be configured in separate PVGs (physical volume groups). This new requirement allows pv_summary to accurately calculate data availability based on physical volume availability, thus including both ACTIVE and INACTIVE volume groups.
If PV links are not configured in separate PVGs, the HA Disk Monitor sees all links to the array as one physical volume. If one link fails, pv_summary will register DOWN, and your package will fail over, even if the other link is still up and data is available."

Huh? but surely that means that my /etc/lvmpvg file needs to look like this:

VG /dev/vgtest
PVG hba1
/dev/dsk/c1t0d0
/dev/dsk/c1t1d0
PVG hba2
/dev/dsk/c2t0d0
/dev/dsk/c2t1d0

Now I can create a LV which is mirrored PVG-strict from c1t0d0 to c2t1d0, but will LVM complain because the alternative link to the first disk is in the same PVG?

Thanks

Duncan

[James R. Ferguson]

Hi Duncan:

I agree that the documentation is confusing. I'd manually setup your /etc/lvmpvg as you describe. Then, try to 'lvcreate' a dummy logical volume. If there are problems with the 'lvmpvg' file's specifications you will have an error reported by 'lvcreate'. If this is the case, I would think your solution would be *not* to declare the alternate links in the first place.

Regards!

...JRF...

[Duncan Edmonstone]

Thanks James, thats what I intend to do.... when I have the hardware available to try it (I don't right now!) Not putting the pvlinks in lvmpvg isn't an option if I want the pv_summary metric to work though...

Cheers

Duncan

[Duncan Edmonstone]

... just pushing this back to the top...

Duncan

[MTSU_SAN]

I am going to be having the exact same problem as soon as my system gets Hardware-ready. I don't suppose the $$ we spend on Service Guard implementation will help, either. If you figure this out, I'd like to know!

P.s. How did you ever install the HA monitors in the first place? They cannot be found on the software depot, or on any application CD I have!

[Duncan Edmonstone]

Jerry,

I'll be waiting for demo kit to become available before I can try this, so don't hold your breath...

If you get to try this before me, be sure to post your experiences here!

With regards to HA Monitors... I'm sure the last time I installed them, they were on the same CD as ServiceGuard. I can't check as I am out of the office for a week now (a weeks training in HP Openview Storage Area Manager - lucky me)

Cheers

Duncan