monitor automountd with ServiceGuard RAC extension

[Thevenet_1]

Hello

I've 2 nodes running Oracle RAC with MCSG.
The 2 nodes also use HA NFS to export Archive log FS.
Each one use Enhanced AutoFS to mount and use Archive log FS.

- NFS packages are monitored and can fail over
- Oracle processes are monitored, and if one is missing the Oracle package is just stopped without failover

I know it's not the only way to configure RAC but do you think I have to also monitor automountd with ServiceGuard (in the same way nfs daemons and Oracle processes are monitored) ?

[Steven E. Protter]

You've gt the RAC extension, which means you should be able to use SG to monitor and failover RAC.

I would think the RAC extension would come with templates to make this pretty easy to do.

If I understand your question (always a doubtful proposition) the summary answer is yes, you can monitor and fail over oracle applications.

SEP

[Thevenet_1]

Steven

Thanks for your answer.

Of course, SGeRAC lets you configure all packages to fail over or not, that's not my probleme :

I wonder if it's advisable to monitor automountd with ServiceGuard since it's used by Oracle through its archive logs ?
This should make NFS mounts more relyable for Oracle ?

[Todd Whitcher]

Hi Thevenet,

Enhanced autofs is a client side application. Its basic function is to mount, monitor activity and umount NFS files. Its a file system manager just like FSTAB but dynamic instead of static. Since it doesnt register as a server side application like other NFS related programs, nfs, lockd etc. its not monitored in the nfs.mon template that comes w/ the HA nfs toolkit. Its not "rpc pingable" like the other NFS related applications.

There is a recomendation when running Enhanced Autofs in a SG env.

From the release notes:
NFS loopback mount - By default, Enhanced AutoFS uses LOFS mounts for locally mounted file systems. Enhanced AutoFS provides an option to allow loopback NFS mounts for the local mount. You need to use this option in the High Availability NFS environments

This helps during package failover.

http://www.docs.hp.com/en/netcom.html#NFS%20Services

Hope this helps.

Todd

[Thevenet_1]

Hi Todd

Thanks a lot, it helps !!!

just let me be sure of it :

- even if automountd is acting like a
logical server (for mount requests coming
from Oracle), I should not monitor it ?

- I've tried to kill automountd for test,
the result is old nfs mounts stay ok,
whereas new mount requests hang

- here Oracle, will do new mount requests to
backup the database

[Todd Whitcher · Accepted solution]

Hi Thevenet,

even if automountd is acting like a
logical server (for mount requests coming
from Oracle), I should not monitor it ?

Correct, there is no template for this. You could customize something via a script if you wanted to check for the IP of automount but there isnt any type of call you could send it to see if its responding. I've not run into any customers setting this up. Typically if Automounter hangs or fails its because the resource its attempting to mount is experienceing a problem (NFS Server not Responging etc). If that service is a NFS-package IP address then Automounter should recover once the package is restarted on the secondary node.


- I've tried to kill automountd for test,
the result is old nfs mounts stay ok,
whereas new mount requests hang

Yes, I would expect that to happen. Automounter is responsible for managing any of the NFS mounts you have in your maps, if it goes away no new mounts will occur. You may even experience issues w/ your currently mounted NFS files. In generally killing the PID for automounter should be avoided, but I understand your testing this out.

- here Oracle, will do new mount requests to
backup the database

This should work fine, use the NFS loopback mount option on your SG nodes that are automounting to the NFS package, thats required for enhanced autofs in a HA environment.

-L Force all mounts to the local host to be NFS mounts
instead of the default LOFS mounts. This is
necessary for highly available NFS mounts.

I hope this helps,

Todd

[Thevenet_1]

Hi Todd

Thank you very much for all your explanations