need advice

[Stan_17]

forgot to add the following changes in #5.

i would also change service name/commands in pkg1.cntl files

SERVICE_NAME[0]="PROD2_mon"
SERVICE_CMD[0]="/etc/cmcluster/pkg1/PROD2.sh monitor"
SERVICE_RESTART[0]=""

SERVICE_NAME[1]="PROD3_mon"
SERVICE_CMD[1]="/etc/cmcluster/pkg1/PROD3.sh monitor"
SERVICE_RESTART[1]=""

SERVICE_NAME[2]="PROD4_mon"
SERVICE_CMD[2]="/etc/cmcluster/pkg1/PROD4.sh monitor"
SERVICE_RESTART[2]=""

[John Poff · Accepted solution]

Hi Stan,

It looks like you've got it all right. You'll have one package which handles three databases, which sounds like what you want. You won't need to do the cmmodpkg in step 2 if you are halting the package with a cmhaltpkg command in step 3. In step 5, you are right about having to rcp the .cntl file and the shell script to the other node.

One trick I like to use when I'm doing package work like this is to comment out the start and stop scripts in the user defined sections, and then test starting and stopping the package on each node. That will let you see if the filesystems are mounted correctly without having to worry about the database starting and stopping. Then, you can uncommment the start and stop scripts in the .cntl and test it again, letting it start and stop the database.

JP

[Ashwani Kashyap]

Hi Stan ,

YOu got everything right . In step 5 , yes you will have to rcp them to all nodes in the cluster .

You know your environment best , but I would put the three databases in three different packages , so that I can bring them down or up using package halt or run commands without affecting the other databases .

[Charles McCary]

Stan,

Hi - we run similarly with multiple oracle db's and sg.

One suggestion, you may want to have individual packages to handle the db's so that you can bring them up and down individually with no effect on the other db's. Just a thought since this seemed to cause you some issues in the past...

thanks,

c

[Stan_17]

Thanks john, ashwani and charles for you comments / tips. i really appreciate them.

john, since i've 'global package switching' enabled for pkg1, when i halt them won't the package switch over to a 'configured node'. that's why i disabled switching first and then halt. do i miss anything here. your tip is very useful and will make use of them. thanks!

ashwani/charles, thanks for your suggestion. i really thought about it, but the reason for me to lean towards 3 databases in a single pacakage is that, all 3 databases are using the same volume group (vg_oracledb) in the disk array and when it comes to maintenance, i can always create 'NoMonitor_' file in /etc/cmcluster/pkg1/ dir, so that it doesn't affect other databases. do i miss to forsee anything else ?

[John Poff]

Hi Stan,

You don't have to worry about the global switching because you are halting the package. The switching only comes into play if the package fails, but since you are halting the package it won't be considered a failure.

JP

[Stan_17]

Hi John,

i got it. thanks.

stan