HACMP config change

[madra]

I've recently took over a system whereby one of the nodes has been migrated to a new environment and thus redundant on the cluster under my support.
The cluster was basically:
Node A, App A, Failover to Node B
Node B, App B, Failover to Node A

The change I want to make it to prevent Node B failing over to Node A, but leave Node B as backup for Node A.

I want to keep the change as simple as possible, so was thinking of just changing the Failover Policy to 'Bring Offline'

Here is the current topolgy, will the change work.
--------------------------------------------
Resource Group sd_nodea_res
Startup Policy Online On Home Node Only
Fallover Policy Fallover To Next Priority Node In The List
Fallback Policy Never Fallback
Participating Nodes nodea nodeb
Service IP Label nodea_svc

Resource Group sd_nodeb_res
Startup Policy Online On Home Node Only
Fallover Policy Fallover To Next Priority Node In The List
Fallback Policy Never Fallback
Participating Nodes nodeb nodea
Service IP Label nodeb_svc


---------------------------------------------



thanks

Madra

 

[MoreFeo]

Well, not sure but I think that you could remove sd_nodeb_res from cluster, and let it run on node B but not managed by HACMP, with the resources (IP, VG, etc...) configured from AIX, and the start script in inittab.

This would let AppA on node A with failover on node B, and AppB running on node B on a stand-alone way.

 

[khalidaaa]

What you said MoreFeo makes since but I would rather make both application under the HACMP cluster control as this will make it easier for starting and stopping scripts and, if any, for monitoring to take place.

madra, changing the Failover Policy for AppB to 'Bring Offline' is all you need! You have to test all possibilities though!

Good Luck.

Regards,
Khalid

 

[sbix]

The simpliest thing.
You should modify the application server scripts for the B application.
The start script should check the hostname and activate the application B only if it is started on node B

 

[khalidaaa]

sbix,

What you said looks simple but how would this help on failover? I guess madra would need to bring Offline AppB better!

After all, there would be some scripting to be done.

Regards,
Khalid

 

[sbix]

Khalid,
madra said "I wish application A can failover, application B not, but I still wish application B ran by the HACMP".
So, in a failover case all the resources normally owned by Node B will migrate to Node A but application B will not run.
This could be a problem only for some specific security reasons (something like "nobody's is protecting Node A while it is mounting Node B Ip address")

 

[khalidaaa]

sbix,

You are right but what will happen in case Node A failover to Node B? Then both application will be running with the suggestion you gave right?

madra wants "to prevent Node B failing over to Node A, but leave Node B as backup for Node A."

All of that can be controlled through brining Node B offline on failover.

Regards,
Khalid

 

[sbix]

Khalid, I said to change ONLY the scripts of the "B" Application... so, in case of FailOver, Application A will continue to behave as usual.
I do agree with you about the Bring Offline settings, which is the "manual suggest"... but sometimes the CED management people forget something, you know there's ALWAYS an operator thinking he's the smartest in the bin

 

[madra]

thanks for your replies.

I'm thinking of going with the script change, that is, creating dummy versions of those in place.

Again thanks.

 

[madra]

As mentioned, I'll be modifying the scripts not to start the application.

Also, I'm thinking of removing the volume group from the resource group, so as not to vary it on or mount the filesystems.

Finally, since my first post (and days holidays), I've come across 'Bring Offline' only being valid with the start policies - 'Online on First Available' and 'Online on All Available'; so I'm guessing I can discount this


Again thanks

 

[ranjit]

Madra,

Not sure if you did get a solution to the initial post. If I understand correctly you have two resource groups (rgA and rgB).

rgA - is to be capable of failover to from NodeA -> NodeB
rgB - should remain on NodeB and never move

The easiest way to achieve this is to change the "Participating Nodes:" list for each resource group to:
sd_nodea_res: nodea nodeb
sd_nodeb_res: nodeb

When sync'g the cluster there will be warning message for RG sd_nodeb_res since it only has 1 node in its "Participating Nodes" list. This can however be ignored. This configuration should achieve your desired fail-over objective.

 

[madra]

I've been out of action for a while with regards this query.

So I'd like to thank you all for your help.