Pros and cons of getting MCSG to monitor Oracle processes?

[zhaogui]

Can anybody tell me what's the pros and cons of having MCSG monitoring Oracle processes?

Thanks,

[Wodisch]

Hi,

I am not certain to understand your question as you intended it...

Are you talking about using MC/SG at all or not to verify your instance is running?

or

Are you asking wether checking for processes in an MC/SG's service is the way to go?

For the first, you could spend a lot of money and go for the "Oracle Parallel Server", i.e. an instance concurrently running on multiple computers - if one fails the survivors continue.

For the second, I do not believe that ONLY checking wether the background annd listener processes are running is sufficient!
"ps" is rather runtime-expensive, and you may have a problem even though all processes are up and running (e.g. filled off-line redo-logfile filesystem).
In my opinion you have to issue real SQL-requests (thruogh the listener) and to verify the result!

Just my $0.02,
Wodisch

[zhaogui]

My question is to ask whether it is good or bad to configure MCSG, so that it can monitor Oracle daemons, such as Oracle smon,pmon,lgwr,dbwr,listener etc. In case if any of these processes die MCSG will try to restart this daemon and if still ailed then MCSG will failover this Oracle Application to adoptive server.
If you run ps -ef|grep PKG then you will see
/usr/bin/sh /etc/cmcluster/PKG/PKG.sh monitor is running ,which is monitoring those Oracle daemons.

[G. Vrijhoeven]

Hi,

We have a 6 node MC/SG cluster running and several packages run oracle databases. We do not use service monitoring scripts. We use ITO scripts and when a process fails an alert is generated so a DBA can log in an check out the problem, fix it and restart oracle. Oracle processes do not stop or hang without a reason so if one fails there must be something wrong.

Hope this will help.

Gideon

[Alexander M. Ermes]

Hi there.
I would not suggest to keep Oracle processes under cluster control. The database may go down for some reason and the cluster will try to restart it over and over again.
Before that a dba should have a close look and decide, what to do.
We had to do the same decision and we do not use cluster control for Oracle.
We only get messages, if databases do not work properly. That should be enough.
Rgds
Alexander M. Ermes

[Kenny Chau]

Please go to this link for information:

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0x40908cc5e03fd6118fff0090279cd0f9,00.html

Regards,
Kenny.

[Jim Carter]

Hello,
There have been a few suggestions regarding not monitoring the Oracle processes with MC/SG, but I think you need to look at your business requirements AND at the reliabiliity of the Oracle instances you are running. First, you probably have this high availability system in place to provide end-user service to mee certain Service Level Agreements with your customers, how does leaving Oracle down until a DBA can look into a problem at 02:00 in the morning impact your SLA? Second, these monitor processes are very stable and only die if someone accidently kills them, there is an o/s problem with servicing them, there is a database problem, or there is a hardware/network problem. In that case, you can have MC/SG attempt to restart Oracle on the current node before switching to another node, and you can decide how many times you want MC/SG to try a restart. This is all completely configurable. If there was a problem with the database instance, and it is recoverable, Oracle 8.1.7 and above usually does so. If it wasn't recoverable, what damage will be done by trying to restart anyway?

I suggest you review your "mission requirements" to determine the answer to you question, because each company is different. I know of a lot of companies however who do monitor these processes; HP provides the monitor program in our Enterprise Toolkit option to MC/ServiceGuard.

Also, talk to your local HP Account Support Enginner regarding any further details you might need. They can help answer questions like this by working through the business issues with you.

Good luck...

[Bart Paulusse]

If Oracle goes down, I would like to have the opportunity to find out what's wrong.
In our environment MCSG switches in case of hardware defects, NOT when Oracle goes down.
We've already experienced a situation where starting Oracle resulted in a restore session that could have been prevented by NOT restarting Oracle but first fixing the 'defect'

a restore is even worse to your SLAs...