Service Guard and OEM

[Shane Russell]

Hi,

I have three Oracle databases running in my 2 node cluster. Each database has its own package. I want to add an OEM agent to each package.

I have added a new service to the .conf file. I have added a new file system and a start/stop/monitor section to the .cntl file.

I can start & stop it fine but the monitoring fails.

Does anyone use OEM to monitor their Oracle databases in a cluster,

thanks in advance
Shane

[RAC_1]

error messages? Try enabling debugging for cntl files and check.

[Shane Russell]

Here is the error from the log file

May 4 14:27:44 - Node "waldorf": Starting service WLRCONP using
"/etc/cmcluster/wlrconp/toolkit.sh monitor"
May 4 14:27:44 - Node "waldorf": Starting service WLR_OEM using
"/etc/cmcluster/wlrconp/toolkit_oem.sh monitor"
cmrunserv : Service name WLR_OEM does not exist.
ERROR: Function start_services
ERROR: Failed to start service WLR_OEM
May 4 14:27:44 - Node "waldorf": Halting service WLRCONP
May 4 14:27:45 - Node "waldorf": Halting service WLR_OEM
cmhaltserv : Service name WLR_OEM is not running.
May 4 14:27:45 - Node "waldorf"
*** /etc/cmcluster/wlrconp/wlr_oem.sh called with stop argument. ***

May 4 14:27:45 - Node "waldorf"
: Aborting OEM Agent at Thu May 4 14:27:45 BST 2006
In the OEM stop function ...

[RAC_1]

Did you check your config. Looks like config files does not such a service name.
Additionally you can also put set -vx in /etc/cmcluster/wlrconp/toolkit.sh and check.

[Stephen Doud]

To add a Service to a package, the package configuration file has to have a set of parameters describing the service:

SERVICE_NAME
SERVICE_FAIL_FAST_ENABLED
SERVICE_HALT_TIMEOUT

If you added them to the configuration file, you have to halt the package and then cmapplyconf the configuration file so that the cluster binary file has a reference for the service name. Use cmviewconf to see what is in the cluster binary.

If the cluster binary has a reference for the new service, then make certain that you add a set of these parameters customized in the package control script:

SERVICE_NAME[0]=""
SERVICE_CMD[0]=""
SERVICE_RESTART[0]=""

If you previously defined a service in the script, don't forget to increment the index number in brackets on this new service definition.

[melvyn burnard]

Well you have supplied you r own answer:
cmrunserv : Service name WLR_OEM does not exist.

You have put the service commands in the control script file, but not configured this service into your package in the package configuration file.

[Greg Vaidman]

I seem to recall a customer I was at last year that was doing this. The problem was that OEM did not recognize the floating IP/cluster name as valid in its' configuration file. So they had to have the "real" hostname in the OEM config file on each system. If this doesn't help, let me know and I'll dig out my notes...

This was 9i, not sure if the same issues exist with 10g.

Good luck,
Greg.

[Christian Schulze]

I am running 10GR2 in a Serviceguard-Cluster. Oracle told me, that this delivered OEM (started via: emctl start dbconsole) is not supported to run in an MC-Serviceguard Environment.
I even played around with editing several config-Files to get this thing running with the Package-IP, but no success.

Christian

[Greg Vaidman]

Christian,

Good to know that Oracle has not corrected this issue ;^{

Did you get it to work? As I mentioned, we got this to work on 9i by keeping the "real" hostnames in the OEM config files on each of the cluster nodes.

--Greg

[Christian Schulze]

Hi,

I played around with creating symlinks in ORACLE_HOME for each physical nodename linked to my package_name. I also greped through most of the files in the config dir of OEM and changed the names, but that did not work perfectly either. (And is not supported by Oracle!)
The thing is, that OEM always reports the DB and the listener to be down, even if it is not.

But most other things are working.

We already have a EM-Grid-Control (10GR1) running on a Windowsbox. Unfortunately this Version of Gridcontrol does not support 10GR2; the Windows-Version for 10GR2 should be available after summer 2006.

Christian