SAP Startup problem

[Suraj_2]

Hi All

We are facing some problem to start SAP packages in cluster environemnt.

Our Scenario is as follows

Cluster Nodes

PRD ( 10.0.2.251)
DEV (10.0.2.252)

Cluster Package Details
Package Name: DBCI
Package IP : 10.0.2.250

There are two SAP R3 instaces are running. One is in DEV , our development server, other one is PRD , production Server.

The cluster package by default is running on PRD and SAP R3 is also running w/o any problem.
But when PRD goes down, we will be stoping the SAP R3 instance in DEV and un mouting the file systems, then the Package(DBCI) will be started on DEV. In this case we are able to start the packge and hence the file systems are mouted on DEV. But we are unable to start the SAP package.

We are getting the following problems.
Oracle in not starting.Even listener service is also not starting.

Can anybody give us some sollutions.

Rgds
Suraj

[Steven E. Protter]

Is there an erorr message when oracle tries to fail over?

What do the cluster logs say when the failover occurs. The error messages will lead to the solution.

You'll need to post some configuration information for me to help.

That being said you should make sure things are happening in the right order.

The failover volume group needs to be activated on the second node before any scripts run that start oracle.

SEP

[Suraj_2]

Hi SEP

The Oracle error mesages are attached as follows.

The problem what we observed is after shifting the package and hence file system to DEV,when we login as Oracle Owner, Oracle environemnet variables are not exported.


As per the clustering is concerned file systems are properly shifting to DEV, in case of PRD failure and cluster log says successful starting of Package.

The more details as follows.

Oracle SID in PRD : PRD
Oracle SID in DEV : DEV

Oracle files owner and group in
PRD : oraprd:dba
DEV : oradev:dba

When the file system is shifted to DEV the owner of shifted file system changes to oradev instead of oraprd. Can this be a problem?

Then we want to start the Oracle and hence SAP against the cluster IP. What changes required for this?

If you can give your email id I can send the configuration files and error logs.

Rgds
Suraj

[Rita C Workman]

I start at the bottom and work my way up....
..so...
have you checked resolution:
does pkg name & ip resolve properly on this failover (Dev) node form production pkg ?

Is everything Oracle (binaries, logs, EVERYTHING) within your package ?

Did you confirm that every Oracle process (including the listener) stopped properly on DEV before attempting to start up Prod on this failover server ?

Did you check your /etc/profile and Oracle account .profile to ensure they are accurate for each server ? If your /etc/profile is tailored strictly for the Dev environment, you may need to move the Prod /etc/profile into place before starting up oracle and put/move the Dev /etc/profile elsewhere till Dev comes back up on that box. Quick & dirty, but if your problem is in environment being exported properly it might help.

Just a couple thoughts,
Rita

[Rita C Workman]

rworkman@wvwcc.org

[Geoff Wild]

Is oracle loaclly mounted or part of the package? (better to be part of the package - especial if you are running 2 instances).

I used to do that with Prod and QA...

How I did it - added a
function customer_defined_halt_qa

to the cntl file (as well as the call to that function in the mainline section)...

Function looked like:

{
print "Checking if qapkg is running"

STATE=`cmviewcl | grep qapkg | awk '{ print $2 }'`
if [[ $STATE = "down" ]]
then
echo "QA is down...starting up IPR"
else
if [[ $STATE = "up" ]]
then
QA=`cmviewcl | grep qapkg | awk '{ print $5 }'`
print "QA is running...on the current host: \c"
print ${QA}
if [[ ${QA} = $(/usr/bin/hostname) ]]
then
print "halting qapkg on ${QA}"
/usr/sbin/cmhaltpkg qapkg
fi
fi
fi
}


So that will stop QA (or dev in your case) - but like I said - you need to make oracle part of the package...

Rgds...Geoff

[Stuart Abramson]

It looks like oracle can't connect to the "250" IP address. Make sure that the package IP address shifted over to the DEV machine.

do you use the same port for Oracle listener on both machines? You should probably use two different ports, so there is no confusion...

[Suraj_2]

Hi All

Thanks for the nice replies. In fact we have logged the call with SAP also, but unfortunately SAP is not able to understand our problem.So again as usual ITRC seems great and hope i may get sollution.

Rita :
: Yes, I can resolve the IP and package name.

... Yes..our all oracle files inclusding binaries,db in package.

Oracle is stopped in DEV before starting the PRD package

/etc/profile is not tailor made for specific host.
I will send the logs and config files to you by tomorrow as am now out of office.


Stuart Abramson :

I don't think port is a problem, as the previous running Oracle instance is shutdown, before starting Oracle in DEV and hence the port will be fred automatically.


Waiting for valuables replies.

Rgds
Suraj

[Rita C Workman]

That's fine, but if you can, you might want to try and attach the files here as well.
You have a couple HP MC/SG authorities on this thread with Melvin and Stephen.

Just a suggestion,
Rita

[Philippe Cussigh]

Hello,

As for the moment we have only this listener error it should be important to get :
- Listener.log trace
- Listener.ora file (name given in the listener.log file at startup)
- The proof the listener is well stopped and the port 1537 free.
Use the debug function of SGSAP (touch debug in /etc/cmcluster) and do some manual tests to start listener manually for instance and see with netstat if port is used and if is starts.

best regards

Philippe

[Suraj_2]

Thanks