basic processes

[Ignacio Javier]

Hi everybody:

Do you guys have a list of the principal process that hpux runs that their fail would cause a system crash ?

I want to make a ServiceGuard Package that contains all the critic hpux system processes so if one fails, all the applications fail over to the other node.

Regards

[Steven E. Protter]

Shalom,

Answer: not really.

What happens when a "primary" system process fails is that the entire system crashes and the node goes down, leaving the cluster.

if your serviceguard packages are configured to run on both nodes, basic services can be maintained. OS functions such as NFS, Samba, or apps like oracle, mysql, httpd, tomcat can be started by a serviceguard package control script on the secondary node.

When a system is running there are certain processes that run, they don't need to fail over to a second node of an SG cluster because they are alraedy running there.

Critical user applications can be configured to run on a floating ip address that follows a package from node to node.

The purpose of serviceguard is to provide high availability, not fault tolerance. It was developed because its impossible to desisgn a perfectly reliable system. Statistically speaking having two nodes provide the same functionality lowers the possibility of failure, therefore loss of service to the users to an acceptable level.

SEP

[Ignacio Javier]

Thanks Shalom

My intention is not to start any critical service in the "fail over node". Just to anticipate a situation that would ocasionate a node crash. I mean, imagine tha a "principal" service goes down, but the system does not crash ( i am not sure if this may occur, i guess it could ). The application is running but not working well. For example, not accesing a disk..

What i would like is to anticipate to this situiation a move the application to the other node.

What do you think ?


Regards