Automounter/Auto FS Configuration

Enabling AutoFS on the HA NFS server needs to be done with caution. However, this kind of configuration is a fully supported configuration if one of these two actions are taken before restarting the AutoFS daemon or before performing any stopping or restarting NFS client (via "nfs.client stop/start") or NFS server (via "nfs.server stop/start"):

	WARNING! The NFS-client or NFS-server must NOT be stopped/started a SG package with NFS exports is started!

Repeat the steps in this section for each node of the cluster and for each external application server host.

Logon as root.

Check that the Automounter is active. In /etc/rc.config.d/nfsconf, the section for the autofs configuration must look like:

AUTOMOUNT=1 
AUTO_MASTER="/etc/auto_master" 
AUTOMOUNT_OPTIONS="-f $AUTO_MASTER"
AUTOMOUNTD_OPTIONS= 
AUTOFS=1

Older installations on HP-UX 10.x and installations without autofs require a slightly different syntax for the "old" automounter:

AUTOMOUNT=1 
AUTO_MASTER="/etc/auto_master"
AUTO_OPTIONS="-f $AUTO_MASTER"

Make sure that at least one NFS client daemon and one NFS server daemon is configured to run. This is required for the automounter to work. Check the listed variables in /etc/rc.config.d/nfsconf. They should be specified as greater or equal to one. For example:

NFS_CLIENT=1
NFS_SERVER=1
NUM_NFSD=4
NUM_NFSIOD=4

Add the following line to your /etc/auto_master file:

/- /etc/auto.direct 

Create a file called /etc/auto.direct.

For each directory configured to be mounted below /export in Step Title not available, add a line to this file. For example:

/usr/sap/trans <relocdbci_s>:/export/usr/sap/trans 
/sapmnt/<SID> <relocdbci_s>:/export/sapmnt/<SID> 

For INFORMIX databases, add another entry. For example:

/informix/<SID> <relocdbci_s>:/export/informix/<SID> 

For the two package concept, typically the relocatable IP address of the database package <relocdb_s> is used.

When configuring AUTOFS the automounter map file /etc/auto.direct must NOT be executable. Make sure to set the appropriate permissions of /etc/auto.direct to 644.

Restart the automounter with:

/sbin/init.d/nfs.client stop 

and

/sbin/init.d/nfs.client start 

	NOTE: Never kill the automount process. Always use nfs.client to stop or start it. Never stop the NFS client while the automounter directories are still in use by some processes. If nfs.client stop reports that some filesystems could not be unmounted, the automounter may refuse to handle them after nfs.client start.

After configuring all automounters by performing the above mentioned steps on all hosts, verify that the setup works correctly to this point. Do not continue with the following application dependent sections without doing this! Perform the tests using the debug-switch

Create a file called /etc/cmcluster/debug on all nodes in the cluster. This enables you to start the MC/ServiceGuard packages without running the SAP R/3 specific steps.

Reboot all machines.

Start the cluster and run all packages. This might be done automatically. If everything is working correctly, the log-file(s) in /etc/cmcluster/<SID> will indicate successful package startup.

On the primary host(s) it should be possible to run the database and the Central Instance. Be sure to start the listener as ora<sid> before trying to bring up an ORACLE database. All application servers should work after starting them manually.

Now, that everything works stop all application servers. Stop the Central Instance and the database. Stop the listener if needed. Do not stop the packages yet.