Shutdown Procedures -- MASTER OFF ALL

[gsherry1]

Hello Forum,

Periodically we shut down for maintenance with the following procedure:

1. Broadcast users to exit clients.
2. dssh command: 'MASTER OFF ALL'
3. shutdown DataStage services.
4. Reboot machine.

Doesn't step 3 automatically perform the equivalent functionality as step 2? Does performing step 2 provide a more cleaner shutdown, then skipping directly to step 3?

Thanks,

Greg

[ray.wurlod]

MASTER OFF ALL is user processes (connected clients, running jobs using DataStage repository facilities). The services are the DataStage resource, telnet and RPC services. Not the same at all. I trust you wait for a decent interval between MASTER OFF ALL and shutting down services, as well as making sure no important jobs are running?

[gsherry1]

Hello Ray,

Thank you for your feedback.

I trust you wait for a decent interval between MASTER OFF ALL and shutting down services, as well as making sure no important jobs are running?

I guess I originally had thought that shutting down the service would do it's best to make sure that connections to clients were closed in an manner that would preserve the integrity of the repository objects.

What is a reasonable amount of time between 2. and 3.? I guess I expected that if the command returned without problem, that it would be safe to proceed to step 3.

- Greg

[ray.wurlod]

The right amount of time is "till they're all gone". You can monitor with netstat or ipcs (UNIX commands but, since you have 7.5x2, you have a UNIX shell in MKS Toolkit). A Windows/DOS equivalent of ipcs (only for DataStage) is shrdump. Windows/DOS has netstat available.