S/W fail over

[Lai Nee Shyang_1]

Hi there,

I've a new system implementation with MCSG. It is a 3 node cluster. The special thing about this system is that on one of the machine it has 3 Oracle instances and HP guys propose to have 3 MCSG pkg to fail over the individual Oracle Instances. They plan to implement process monitoring scripts for triggering fail over. I always thot that MCSG protects against HW/NEtwork fail over, never on s/w fail over.

Has anyone out there done something similar?
Any comments please.


Thanks and have a nice day.

Lai
: )

[Alexander M. Ermes · Accepted solution]

Hi there.
Yes, we have something similar here.
We have setup the cluster in case one of the packages goes down on one node, it switches to the other node with the shared disks. Then the Oracle instances run on the second node. It is very useful, if you are in need of a minimum downtime for the users. The package switching takes about 2 minutes or less.
Rgds
Alexander M. Ermes

[Animesh Chakraborty]

Hi,
Offcourse..
It does s/w (package) failover too.
Once in my office someone killed one oracle processes by mistake then the package started on another node successfully.


Thanks
Animesh

[linuxfan]

Hi Lai,

It is very common to have a package defined for oracle instance and in your case have 3 packages defined for 3 instances, what it does is provide flexibility and differentiates each instance and you could(depending on your configuration) run each instance on different node.
This way rollover upgrades become easier, you could upgrade oracle on one node while you failover the package to other node (though this can be done with just one oracle package too).
Not sure if you have seen the doc "Managing MC/SG"
http://docs.hp.com/hpux/pdf/B3936-90045.pdf

-Ramesh

[linuxfan]

Hi Lai,

One more thing, as far as monitoring the oracle databases are concerened, with-in MC/SG the monitoring scripts can be configured to monitor oracle listener as well as individual processes like (smon, pmon etc).

You can actually write your own control/monitoring scripts, but HP also provides the Oracle toolkit($$$) which you can be modify if you want to.

-HTH
Ramesh

[Rainer von Bongartz]

One thing to think about protecting yourself against SW problems with MC/SG: Normally it makes no sense to switch the package over to another node in case of a SW problem. If your nodes are configured in the same way the SW problem will also happen on the next box. So configure MC/SG to re-start your package on the same box and to only re-start it once after a SW error.

I once had a SW problem that my database crashed immediatly after startup and MC/SG restarted it, and then
crash,restart,crash,restart...
about 400 times during a night.


Regards
Rainer

[Printaporn_1]

Actually you just mis understand the wording.
HW should be redundancy like using disk array.
In MCSG there are standby LAN
that can be local fail-over when primary lan fail
package still runing in same machine but standby lan will up to be active lan.
in case of primary node fail the S/W package will fail-over to adoptive node , this is S/W fail-over.

[Lai Nee Shyang_1]

Hi People,

Thanks for your comments. They are really helpful.

Special thanks to Printaporn. You just cleared another miss-understanding I have regarding MCSG definitions.

Thanks to all again.


Lai
: )