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

nfs performance tuning for hp-ux 11.0 and 11i systems page 58

Notes:

Page 58July 22, 2002

Troubleshooting rpc.mountd

• Use rpcinfo(1M) command to “ping” rpc.mountd

• Collect a debug-level rpc.mountd logfile via the

SIGUSR2 toggle mechanism

• Verify that hostname resolution servers (i.e. DNS,

NIS, etc.) are responding and return accurate data

• Collect a network trace of the problem

• Determine if the undocumented rpc.mountd “–X0” option

can safely be used in your environment

rpc.mountd

The rpcinfo(1M) command can “ping” rpc.mountd to quickly determine if the

daemon is at least responding to requests.

One of the best sources of information for troubleshooting rpc.mountd problems is

a debug mountd logfile collected while reproducing the failure. This debug

logging can be toggled on and off by sending the SIGUSR2 signal to the running

mountd process (i.e. kill –17 <mountd pid>). By default, the debug information is

appended to the /var/adm/mountd.log file. In some cases, a network trace is

also needed to fully understand the root cause of rpc.mountd’s non-responsiveness.

A large majority of rpc.mountd issues are caused by hostname resolution problems

(i.e. down DNS/NIS servers). Be sure to verify that hostname resolution is

working.

An rpc.mountd/automount deadlock situation was discovered involving a MOUNT

request of a symbolic link that referenced a loopback NFS mount point managed

by automounter. The fix for this extreme corner-case issue involved adding many

new conditional checks to mountd’s MOUNT function. Launching rpc.mountd with

the undocumented “-X0” option disables these checks.