SAP R3 Mc Service guard

[sreeramam]

Dear All,

We are installing SAP R/3 distributed system with mc service guard.

We have installed CI and DI on different hosts.

Kindly suggest how do we go about configuiring clustere and failover

Waiting for positive reply

Sree Ramam SNV

[Geoff Wild · Accepted solution]

I used to have CI and DB on seperate hosts - had a 4 node cluster - with 2 nodes for DB and 2 nodes for CI....

When I upgraded 2 years ago to 7410's, basis integrated DB/CI on one server - so only 2 nodes required.

For setup, check out this doc:

http://docs.hp.com/cgi-bin/onlinedocs.py?mpn=B7885-90018&service=hpux&path=00/00/20&title=Managing%20ServiceGuard%20Extension%20for%20SAP%2C%20June%202003

And more here:

http://docs.hp.com/hpux/ha/index.html

Rgds...Geoff

[sreeramam]

Very usefull information

Can i have the file system distribution when you had CI and DB on different hosts and on single host.
Tha problem i am facing is when i install central system usr/sap/MKS/SYS/run/exe is having app. related binaries and in DB server in usr/sap/MKS/SYS/run/exe we have all db related file such as sapdba, br* etc

In the event of failure of one node the
usr/sap/MKS/SYS/run/exe is to be mounted but i already have usr/sap/MKS/SYS/run/exe munted then how does it do


That means what i understand is for a distributed system to work in clustered mode i need four nodes and for central system to work we need two nodes.
I assume that all this is possible only if have common storage. (SAN).

But in this scenario one of my nodes is is always idel if i am correect in my understanding.

Now i am really great for you paticence to come through such a long script..

So finally one more thing that i would like you to suggest is

I want distributed system to be running but in the event of db/ci failure my ci/db should take over

Is this possible..

finally i would like to have your mail id and phone no.

I am from india,Ahmedabad

Waiting fro positive reply

[Geoff Wild]

Sree - currently I'm on a HP course - if I get a chance, I'll connect to work from my hotel later - and see if I can post my old scripts...

Just so you know - we didn't monitor for CI/Oracle failures - only Hardware/Network...

Rgds...Geoff

[Geoff Wild]

Sree - check out this thread I posted to earlier this year:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=592985

Rgds...Geoff

[Geoff Wild]

Just noticed that those inks I gave you in my last post was how to handle running a QA and Prod system in same cluster...

Answering your other questions:

"That means what i understand is for a distributed system to work in clustered mode i need four nodes and for central system to work we need two nodes.
I assume that all this is possible only if have common storage. (SAN).

But in this scenario one of my nodes is is always idel if i am correect in my understanding."

That is correct - you would have 4 nodes, 2 dedicated for DB and 2 dedicated for CI - and at any given momemnt - 2 would be idle.

And yes, storage has to be common - between the pairs of nodes.


"I want distributed system to be running but in the event of db/ci failure my ci/db should take over "

If I understand you correctly, it would be very difficult to have the DB running on the CI nodes in a failover...

Here's what has to happen - regardless of configuration...

If the DB was to failover, you need to check that the CI is NOT running - if it is, you need to shutdown CI first, then start DB.

If CI was to failover - no issues...DB can stay up - as it is not dependent on CI...

You will also need a TRANS package - for transports - this is a NFS package - and will have to be running prior to DB and CI...

The reason we went to an integrated DB/CI server was for simplicity - also our DEV and QA landscapes are setup that way.

I highly suggest you do the same - unless the hardware you have can't handle it...

Rgds...Geoff

[sreeramam]

Hi,

I got a query..

In the event of failure do i need to start sap on other node when package is shifted


Will the user be able to sense the down and up of SAP.

Will the changeover take place automatically without Admin involment.

[Geoff Wild]

To answer your last query - are you going to run this on a 2 node or 4 node cluster?

If on 2 node - then you might as well as integrate CI into the DB package - for example, if one of the nodes has a hardware failure, you will have to run both on one node anyways...

In the event of a DB failure, you will have to restart CI...so you will have to have a check in the DB cntl script to check to see if CI package is up - if yes, shut it down, then restart after DB is up...

Users will see the hourglass until you are up - but they shouldn't have to re-connect - as they connnect to your App Servers - which should be up...

For automatic - that's up to you - you will have to script it...

Rgds...Geoff

[Geoff Wild]

Don't know why I didn't think of this before:

http://docs.hp.com/hpux/onlinedocs/B7885-90013/B7885-90013.html

That will do everything you are looking for.


Rgds...Geoff