ONCplus B.11.31.09.01 Release Notes HP-UX 11i v3 (September 2010)

3 Known Problems and Workarounds

This section lists known problems, probable cause, and appropriate workarounds.

Problem

There is a possibility for an NFS client machine to experience a system hang if you are using the

below ONCplus versions:

• ONCplus B.11.31.09

• ONCplus B.11.31.09.01

Cause

The NFS client machine could hang under below conditions/configurations:

• Severe file cache memory pressure due to low file cache configuration.

For example, filecache_min = filecache_max =1%

• When the nfs3_max_threads tunable is set to “0”.

Workaround

The fix for this issue will be available in ONCplus B.11.31.10, which is planned to be released in

September 2010.

Following are the possible workarounds to handle this issue,

NOTE: All these workarounds must be considered to ensure the system hang does not occur.

• Set the below values to the NFS internal variables:

# echo "async_commit_wait_multipler ?W 0xF4240" | adb -w -o /stand/vmunix /dev/kmem

# echo "nfs_awcount_multipler ?W 0x2" | adb -w -o /stand/vmunix /dev/kmem

• Increase the value of the nfs3_max_threads tunable to least 8 using the following command:

# kctune nfs3_max_threads=default

• If you are doing a manual un-mount, ensure that no application is accessing the NFS mount

point at the time of manual un-mount.

For example, if /mnt/nfs is an NFS mounted directory, use the following command to

verify and un-mount the directory.

# fuser –c /mnt/nfs

• If you are using AutoFS to un-mount an NFS file system, then disable automatic un-mount

by setting a high time out value.

For example, set AUTOMOUNT_TIMEOUT to 99999999 in the /etc/default/autofs

file.