NFS Performance Tuning for HP-UX 11.0 and 11i Systems

nfs performance tuning for hp-ux 11.0 and 11i systems page 69

Notes:

Page 69July 22, 2002

How are Automount and AutoFS

different from each other? (part 2)

automount

autofs

kill –9 is safe if necessarykill –9 will hang filesystems

keeps reference timer (for direct

maps only) – avoids attempting

to unmount busy filesystems

doesn’t keep track of which

filesystems are in use – issues

unnecessary unmount requests

maps changes take effect

immediately whenever the

/usr/sbin/automount

command is issued

adding mount points to master

or direct maps requires

automount be killed and

restarted to take effect

Automount AutoFS

Any changes made to the master or direct maps requires automount to be stopped

and restarted for the changes to take effect. AutoFS re-processes its maps

whenever /usr/sbin/automount is issued. Changes take effect immediately.

Automount attempts to unmount filesystems at the specified interval regardless of

whether they are in use or not. AutoFS has the ability to track when filesystems

were last referenced and only attempts to unmount those filesystems that have been

idle for the specified interval. Unfortunately, HP’s current implementation of AutoFS

only supports this feature for filesystems configured in direct maps. Indirect and

hierarchical map entries behave like the old automounter.

Since automount is a user-space daemon that puts placeholders in the VFS layer, if

you kill the automount process with a –9 (SIGKILL) it will forcibly exit without first

removing the VFS placeholders. Now any access to an automount-managed

directory will hang, as the kernel views these entries as belonging to a dead NFS

server. The portions of AutoFS that manage the VFS layer entries are in the kernel,

which makes it “safe” (although not recommended) to stop the user-space

automountd daemon with a SIGKILL if necessary.