Operating System - HP-UX
1834736 Members
2856 Online
110070 Solutions
New Discussion

Re: Serviceguard fail over

 
ckchai
Frequent Advisor

Serviceguard fail over

Hi,

My environment:-

3nodes cluster- HPUX 11.11

1st node - Oracle DB (instance name=CIM).
2nd node- Oracle DB (instance name=space).
3rd node - standby node.

I want to shutdown Oracle DB (space) at 2nd node to do maintenance. I also don't want the Oracle DB (space) at 2nd node to failover to the standby node. What is the correct steps that I should follow?

regards,
ck
9 REPLIES 9
JASH_2
Trusted Contributor

Re: Serviceguard fail over

CK,

If you halt the package manually, it should not fail over to the standby node.

Regards,

JASH
If I can, I will!
Chan 007
Honored Contributor

Re: Serviceguard fail over

Hi CK,

Just do the following

1. On node 2
cmhaltpkg -v

Once you complete the maintenance to start on node 2
cmrunpkg -v

Tip: IF you manually shutdown any package or a node, that will not failover.

Hope this helps

Chan
ckchai
Frequent Advisor

Re: Serviceguard fail over

Sorry. I forgot to tell that, actually i did try to do the cmhaltpkg.
but once I cmhaltpkg, all the oracle DB files located at the quorum or shared disk is unmount.

Chan:- Are you saying that if i do a manual shutdown Oracle (shutdown immediate), the current cluster won't failover the application to the standby node?
Warren_9
Honored Contributor

Re: Serviceguard fail over

Hi CK,

which version of the enterprise toolkit is using?

if you want to have the shared disk mounted, there is couple ways to do it.
1. shutdown the oracle package and then manually activate the VG and mount the FS.

2. there is a bypass flag in the new version of ETK. Touch the flag in the oracle package directory and it will bypass the monitoring of the oracle processes.... then shutdown the oracle to do the maintenance.
remove the flag after startup the oracle.

GOOD LUCK!!
Tiziano Contorno _
Valued Contributor

Re: Serviceguard fail over

It depends on the controls you implemented.

I suggest you (to have a cleaner Service Guard status) to halt the package (cmhaltpkg) and activate manually the volume group/s (vgchange -a e) and mount the filesystems, if you need them mounted.

Once done, unmount fs and deactivate vg.

Regards.
Warren_9
Honored Contributor

Re: Serviceguard fail over

CK,

BTW, as I remember, there is infomation in the master toolkit readme file .

check the /opt/cmcluster/toolkit/oracle/readme...

GOOD LUCK!!
Chan 007
Honored Contributor

Re: Serviceguard fail over

CK,

Check that your oracle DB or applications are not being monitored, you can work without shutting down the package.

Once that cluster package is running, do an oracle shutdown and do your maintenance.

Chan
shardam
Frequent Advisor

Re: Serviceguard fail over

Hi CK,

If the 2nd node will be halted you need to fail over the package which residing from it to another node so the production from the package will not be affected. Much better to fail over the package from 2nd node into primary node.

Best regards,
Alfredo
Ajitkumar Rane
Trusted Contributor

Re: Serviceguard fail over

CK,

If you dont want the package (space) to failover to other node, do a cmmodpkg -d space. and then you can do cmhaltpkg. other wise if you do a cmhaltpkg without disabling the auto failover it will failover to the adoptive node if you just halt the pkg or node.

Rgds,
Ajit
Amidsts difficulties lie opportunities