topic Re: Service process in Operating System - HP-UX

Service process

Andrew R. — Sun, 12 May 2002 22:57:07 GMT

Background:
I have two servers, Both have vg00 and vg01 (2 mirrored drives [HSC 72GB] on each VG, Primary swap lvol2 in vg00 and Secondary fileystem swap in vg01).The Application is running from vg01. Drives are not shared between the servers.
We have XP256 that I could possibly connect it to the fibre cards on both servers if needed.

Objective:
I need to configure the cluster between the two servers to monitor the Application Processes ONLY (No Application Failover)

Q1:
Do I need to share vg01 between the two servers ? Isn't the Lock disk has to be shared between the two servers ?

Q2:
If that's the case, Should I move the swap to vg00 (which is not a good idea)

Any help is very much appreciated.
Thanks

Re: Service process

rajsri — Mon, 13 May 2002 00:35:36 GMT

Q1:
Do I need to share vg01 between the two servers ? Isn't the Lock disk has to be shared between the two servers ?

A. If it is going to be 2 node cluster, then you definately need cluster lock disk, but if you dont want to share vg01, bcs of swap lv , you might create another vg with single lun and share between two systems, which will act as cluster lock disk. As you have XP you can create least possible size LUN and use that LUN for cluster lock disk.

Q2:
If that's the case, Should I move the swap to vg00 (which is not a good idea)
To avoid this create another VG as mentioned in above.

Regards,
Raj

Re: Service process

hari jayaram_1 — Mon, 13 May 2002 19:53:39 GMT

Hi Andrew,

Objective:
I need to configure the cluster between the two servers to monitor the
Application Processes ONLY (No Application Failover)

Q1:
Do I need to share vg01 between the two servers ? Isn't the Lock
disk has to be shared between the two servers ?

Are you running MC service guard or Veritas clustering software. Please advice.

Re: Service process

Duncan Edmonstone — Mon, 13 May 2002 19:59:44 GMT

Andrew,

If you don't want failover of any kind, then why bother clustering at all? Just write some monitoring scripts yourself to monitor for applictaion failure locally. MCSG (or VCS for that matter) just use scripts to handle application monitoring anyway. A few monitor scripts will probably be simpler (and cheaper to boot)

HTH

Duncan

Re: Service process

Andrew R. — Tue, 14 May 2002 00:05:25 GMT

Thanks for your responses,

I'm using MCSG /OS 11i
The reason I need to install Service Guard is to cover the Network failover, I have 4 Lan cards and 4 Standby Lan cards. The second reason is If I write some scripts it will not protect me from HW failure. I want to use the Service function in SG. I know that it sounds over killing but do I have alternative ?

Re: Service process

Duncan Edmonstone — Tue, 14 May 2002 07:20:25 GMT

1. If your just using MCSG for the network fail-over then set the two boxes up as two seperate single node clusters. Then there is no requirement for any shared disk at all.

2. If you are not failing over, the only hardware protection that MCSG will give you is the afore-mentioned network card fail-over.

Have you looked at Auto-Port-Aggregation (APA) - this will provide the network resilience without all the other bits & pieces of MCSG.

HTH

Duncan

Re: Service process

Duncan Edmonstone — Tue, 14 May 2002 07:30:41 GMT

I meant to put this link in the previous post:

http://www.hp.com/products1/unixserverconnectivity/software/overview_features.html

Of course you can only use APA if your network switch supports it.

HTH

Duncan

Re: Service process

Andrew R. — Tue, 14 May 2002 13:08:07 GMT

Thanks Duncan,

Our switches doesn't support APA.
I believe if you have the process being monitored under MCSG and system A paniced these processes will start on system B ... am I correct ?