Correlation Between application and Packages

[Karthik Narayanan_3]

Hi,
Is there any way to correlate the application and packages reported by the native SG agent.
If I create a Package name "Test" , the application named "Test" is automatically created? Is this always the case?
Can a Package contain more than one application?

Please advice.

Thanks
-- Karthik

[Sanjay_6]

Hi,

there is no correlation between the package name and the application name. We keep the name common most of the time so that just by looking at the package name, it can be said what application is running on that box. But you can name them anything you want.

A SG package can have more than one application running.

Hope this helps.

Regds

[Jeff Schussele]

Hi karthik,

Certainly, a good example would be Oracle - the application itself & it's listener are 2 separate apps.
I think the only limit to the # of apps in a pkg would be practicality or resources.
Rgds,
Jeff

[A. Clay Stephenson]

A package could consist of any number of applications that share a common IP address and resources (e.g. filesystems) BUT life is generally simpler when there is a one-to-one relationship between packages and applications.
Note that in this context 'application' might refer to many individual processes.

[Karthik Narayanan_3]

Hi,
Is there any way then to manage applications separately using what the SG agent reports.

Thanks
Karthik

[A. Clay Stephenson]

I'm not quite sure what you are looking for. For example, it is possible that your package might contain 3 separate Oracle instances. You could do svrmgrl scripts for each of them. You can also use svrmgrl to manage completely outside of MC/SG. However, in general, it is by far best to let the package manage everything as a unit. Consider, how complicated you are making failover if everything is not done as a unit.

[Karthik Narayanan_3]

Hi,
Thanks for the reply. I would like to clarify my question.
Is there any way to 'monitor' the package and its associated applications using what the SG agent.

Thanks
Karthik

[Duncan Edmonstone]

If by 'monitor' you mean 'make sure the application is still running OK' then yes, definately - thats one of the standard features of ServiceGuard - each package can have a number of 'services' associated with it - the purpose of these services being to monitor the state of the application - one 'service' comes built in - it monitors the IP subnet that the package (ergo application) uses for communicating with its users (clients). Other services have to be hand-crafted, but this is usually a relatively straightforward task. When a package starts, the ServiceGuard also starts all services associated with the package. When a service exits for any reason ServiceGuard can either choose to restart the service, or fail-over the package (this is configurable). In this way it is very easy to write application monitors which take the form of:

while (true)
do
if (application processes are no longer in process table) then break
done
exit

And these can, of course be written in the language of your own choice (shell/perl etc.)

To learn more about this I would strongly advise you to read the ServiceGuard manual here:

http://docs.hp.com/hpux/onlinedocs/B3936-90065/B3936-90065.html

HTH

Duncan