Need Example of SERVICE_CMD for Oracle in SG 11.15

[Greg OBarr]

In migrating from SG 11.09 to 11.15, there appear to be some differences in the way the cluster starts and monitors Oracle. I'm running Oracle 10g in SG 11.15, and I was running Oracle 8.1.7.4 in SG 11.09.

The 11.09 package control script has this for the SERVICE_CMD:

SERVICE_CMD[0]="/etc/cmcluster/PROD/PROD.sh monitor"

The new name and location of my package control script in SG 11.15 is:

/etc/cmcluster/pkg/PROD10G/PROD10g_pkg.sh

So, I'm thinking it should be:

SERVICE_CMD="/etc/cmcluster/pkg/PROD10G/PROD10g_pkg.sh monitor"

Is this correct?

[Greg OBarr]

Correction:

It appears that it would be best to keep the name "haoracle.sh" for the package script, so I think the command should be:

SERVICE_CMD[0]="/etc/cmcluster/pkg/PROD10G/haoracle.sh monitor"

Comments?

[melvyn burnard]

I think your second choice is better. Are you using the latest Oracle Toolkit?

And what command do you use to start your application, i.e. the user_defined_run_command section of the pkg control script?

[Eric Antunes]

Hi Greg,

I don't work with service guard but it seems ok to me.

You may need to define some other variables (I think you'' find some template files somewhere in the SG path...):

SERVICE_NAME[0]=

SERVICE_RESTART[0]=""
SERVICE_RESTART[0]="-r 2" # Will restart the service twice
SERVICE_RESTART[0]="-R" # Will restart the service an infinite number of times.

Regards,

Eric

[Eric Antunes]

Hi,

Read also this Note about SG and IA's:

http://metalink.oracle.com/metalink/plsql/ml2_documents.showDocument?p_database_id=NOT&p_id=189401.1

Eric

[Greg OBarr]

Melvin,

Start script is:

/etc/cmcluster/pkg/PROD10G/haoracle.sh start

test_return 0

Halt Script is:

/etc/cmcluster/pkg/PROD10G/haoracle.sh stop

test_return 0

Startup and stop seems to be working, but packaage startup seems to be failing. The log file looks like it's detecting that an oracle process has stopped and is shutting everything down. Looks like there are additional processes to monitor in 10g, and I'm trying to get that all straightened out.

When I start 10g manually, I see the following processes with no load on the databse:

ora_lgwr_PROD10G
ora_ckpt_PROD10G
ora_reco_PROD10G
ora_cjq0_PROD10G
ora_pmon_PROD10G
ora_s000_PROD10G
ora_mmnl_PROD10G
ora_p000_PROD10G
ora_d000_PROD10G
ora_q000_PROD10G
ora_p001_PROD10G
ora_qmnc_PROD10G
ora_smon_PROD10G
ora_mmon_PROD10G
ora_j000_PROD10G
ora_mman_PROD10G
ora_dbw0_PROD10G

Should I monitor all of these, or just the ones defined in the template? It appears that the template is designed for Oracle 9i rather than 10g, and there are additional processes in 10g.

Thx.

[Greg OBarr]

Eric,

In 10g, I'm using the "emctl start dbconsole" command to run the enterprise manager web interface rather than the old "intelligent agent" that came with 9i and earlier.

-greg

[melvyn burnard]

Then you have the correct service command as per your example. If you are using the older scripts, then yes these will not be aware of the 10g changes (if any).
You should install the newer Oracle Toolkit scripts which cater for 10g.
Take a look at the release Notes:
http://docs.hp.com/en/T1909-90019/T1909-90019.pdf

[Greg OBarr]

You're right, Melvin. I have the B.02.00 version of the Enterprise Cluster Master Toolkit installed. I need B.02.20. Should this be on one of the support CDs? Or can I download it?

[melvyn burnard]

This should be on your application CD's. It is not downloadable unless you pay for it online.

[Greg OBarr]

Thanks,

I was able to download it from the Software Update Manager site...

-greg