Thanks to the following for their help/assurance :
Frank Wortner
Alan Davis
Allan Simeone
Cherie Willoughby
Actually, I basically had it right. It will work as desired. I also
received a few good tips too.
Here's a few of the snippets...
################################
Relax! We've got a similar set up here. It's Informix rather than Oracle,
but the idea is the same. The service start action scripts run after the
disks are mounted and the stop action scripts before the disks are
unmounted. Works just fine.
################################
As long as you follow the rules, the service start/stop scripts will "do
the right thing". There are scripts that are internal to the service start
and stop that handle making sure that the storage is available, the exports
are made, etc. The parts that you have control over are starting and
stopping the applications and in some cases custom mods to the network
handling.
Some of the key issues are :
Start all applications that use disks in the service
in one "service" start/stop script
Don't access storage from the service application that
isn't included in the services storage list
Understand fully the actions taken on failover for
various failures. In particular, network failures
don't always have the effects on failover that you
expect
Networks must be defined before configuring TCR
Never access service disks by anything but the nfs mount point
fuser is your friend
all scripts must return 0 or non-0 codes for success/fail
###################################
You got the logic right. All of the disks come online before the start
script executes and all of the processes assosiated with the disks need to
cease before the disks can unmount and the service can stop. The action
scripts wouldn't make much functional sense if they didn't work this way.
However, think about it a little differently. The disk service will not
start until the start script successfully executes and the service will not
stop unless the stop script successfully executes. The scripts have much
impact and if there is a problem with those scripts, those services will
not stop/start at all.
One thing to keep in mind, I had to add a kill function to my stop scripts
because if people are in any of the disk areas, the disks can't unmount and
hence the service can't stop. You might want to include this in your stop
scripts too if any users/admins might end up in those filesystems.
######################################
-----Original Message-----
From: Wayne Sweatt [SMTP:sweatt_at_dps.state.nm.us]
Sent: Wednesday, September 29, 1999 12:11 PM
To: 'DU Mail List'
Subject: Simple TruCLuster ASE Action Script question...
Yes, I just got back from a class, but now that I sit down
to set up my own real-life ASE service, I hit a conceptual wall:
DU 4.0D - TruCluster Available Server V 1.5
Database(Oracle) entirely on shared disk (datafiles, control files, redo
logs) on 4 AdvFS partitions.
Database Server (Oracle 8.1.5) on each member local disk.
When do the action scripts run ? Specifically the start and stop scripts.
The training guide does not state WHEN they run relative to the service
itself.
I planned to run an Oracle Database "service" around a disk service with
start/stop scripts.
I hope that the start script runs AFTER the disk service starts and
the stop script runs before the disk service stopping.
You have to have the shared disks mounted BEFORE you try and bring up the
database/listener.
You have to unmount the disks again AFTER you stop the database/listener
The class demos indicated this... but I want to make sure first.
Thanks,
***********************
Wayne Sweatt
Principal Software Analyst
Litton / PRC
505.827.9288
***********************
Received on Wed Sep 29 1999 - 21:44:48 NZST