Managing Serviceguard Extension for SAP on Linux (IA64 Integrity and x86_64), April 2009

To configure the NFS automounter the file system /usr/sap/trans will again be used as an example.

The goal is to access /usr/sap/trans transparently from all cluster nodes. The initial approach for

configuring the automounter on Linux:

1. As in the static NFS case from above: On the cluster node acting as the NFS server the

/usr/sap/trans volumes are mounted under an /export directory and exported to the NFS clients.

Enable a virtual host /nfsreloc (relocatable IP address) from where the NFS clients can mount file

systems. The virtual host has the advantage to it is always enabled on the cluster node where the NFS

directories are being exported from and the NFS clients only have to mount from this name.

2. On the NFS clients (regardless if cluster node or remote system) auto mount the NFS directories under

an /import directory. The /import directory will be used as a pseudo mountpoint because Linux

does not yet support direct map mount points when a parent mount point exists. For example, if the

automounter map /usr/sap/trans nfsreloc:/export/usr/sap/trans is used it will fail since

target directory /usr of /usr/sap/trans exists on a Linux system. This is because the Linux

automounter wants to remove any files in target directory /usr before executing. So since we are

using the automounter Linux the only way it will work is to use a pseudo mountpoint and create symbolic

links from SAP directories pointing to the pseudo mount points. The /import directory will just serve

as dummy place holder.

3. The configuration of automounter is based on two files /etc/auto.master and a mount specific file

/etc/auto.import (also called an automounter map file).

An example of automounter map file in /etc/auto.master:

/import /etc/auto.import nfsvers=3

The automounter map file itself /etc/auto.import contains the following line:

trans -fstype=nfs,rw,udp,nosymlink nfsreloc: /export/usr/sap/trans

For each line in /etc/auto.master a separate automount process is started. If a user or a process

tries to access any files or directories in /import (for example /import/usr/sap/trans), the

automounter will read the automounter map file /etc/auto.import and find a description of how

file systems below the /import mount point of the map are to be mounted. The basic format of a line

in such maps is: key [-options] location. The automounter then tries to generate a unique key

from filename /import/usr/sap/trans that will map to a line in /etc/auto.import. In this

example trans is a unique key.

NOTE: The names of the sub-directories in /import have to be chosen so that they map into a unique

key / name pair. In the case of example file systems /usr/sap/trans and /usr/sap/tmp the

key sap cannot be chosen as it would not create a unique mapping. Therefore keys trans and tmp have

to be chosen.

4. Create a symbolic link that points from the original SAP directory names to automount NFS place holders

in holders in /import

ln -sf /import/trans /usr/sap/trans

Summary of automount file systems

The following is a summary of file systems that will be mounted with the automounter: