Disabling failover when the system get rebooted manually - Possible options

[Pandi]

Hi,

We are going to perform a scheduled reboot in the primary node which is serving only one service group.

We do not want to switch over the SG to the backup node while the node gets rebooted.

How to do that. I heard that while rebooting the primary ndoe vcs 5.0 will invoke hastop -local -evacuate ..so i tried to disable the autoevacuate in the SG.

I set autoevacuate to 255 ( thought anyvalue greater than 0 will disable it). But it didn't worked . the SG failed over to next node while the server gets rebooted.

2. Disable the autofailover for the SG.

3.when i booked a case with symantec they suggested to freeze  the sg temporarily or permamnently.

but as far as i know,  Temporary freeze willl work till the had stops. so in that case the method will fail..so a permanent freeze of SG will help ?

4. The symantec engineer asked me to offline the sg, freeze it. and then do the reboot...- will it work ?

What is the VCS behaviour when i reboot the server manually..how to keep the SG in the same primary node before and after the reboot ?

It should happen only for manual reboot. if a system crashes the vcs failover should work....

multiple options..got confused :)

[eu22106]

 I would indeed offline the Service group, freeze it and then reboot.
We experienced some diffuculties when rebooting (windows) if the SG group was not brought offline. We had to import the disks, SCI reservations, rescan. You know the lot.
Which makes it more difficult.

Why not just evacuate and the again on the second node. Let VCS do it's work?

Ivo

[Gaurav_S]

Hello,

Well, if you just want to stop the failover for manual reboots, then yes persistent freeze is your answer.... so offline group, persistent freeze it for cluster & then reboot your box.... Also make sure it is persistent freeze only.....

In case you tune any parameters like "Autofailover" or "autoevacuate" then cluster behaviour will change for both manual reboot & system crashes, so this will not be suited to your requirement right now....

Hope this clears your confusion....  :)


Gaurav

[Pandi]

Hi

I did the above method, but the cluster didn't start after reboot ( the sg autodisabled). So i did a hastart and unfreeze the sg.

Once i unfroze the sg , the sg became online.

[Gaurav_S]

Hello Pandi,

Little confused with your statement...

If SG was autodisabled (which can be expected since had was down on one node), why it was needed to do hastart , cluster was already running !! & since had was starting, autodisabled flag should have got cleared automatically.....

If I understand, ideally things would have came to same frozen state once node was rebooted.... quite possible that you didn't waited enough till cluster was resuming ?

Gaurav

[Pandi]

thanks gaurav,

yes your correct, only if had is running it can autodisable the sg.

as far as i know , i waited long enough to start the had.

i am waiting for another downtime. i will do the same method and let you know the results. :)