Operating System - HP-UX
1833541 Members
3200 Online
110061 Solutions
New Discussion

HPUX 10.20 - long delays when attempting to access file on unavailable automounted file system

 
Matt Banek
New Member

HPUX 10.20 - long delays when attempting to access file on unavailable automounted file system

Is there any way to prevent automountd from waiting upwards of 1
minute when a file access is attempted on an unavailable automounted
file system? I have not changed the default NFS timeouts/retry counts
in the kernel.

My /etc/rc.config.d/nfsconf is the following:
-- BEGIN --
NFS_CLIENT=1
NFS_SERVER=1
NUM_NFSD=16
NUM_NFSIOD=8
PCNFS_SERVER=0
LOCKD_OPTIONS=""
STATD_OPTIONS=""
MOUNTD_OPTIONS=""
AUTOMOUNT=1
AUTO_MASTER="/etc/auto_master"
AUTO_OPTIONS="-f $AUTO_MASTER"
START_MOUNTD=1
MOUNTD_VER=3
AUTOMOUNT_OPTIONS=
AUTOMOUNTD_OPTIONS=
AUTOFS=1
MOUNT_VER=3
-- END --

My /etc/automount_master is the following:
-- BEGIN --
/net -hosts -soft
-- END --

I used "trace" to monitor the automountd process while the following
command is executed: "ls /net/"
The trace output shows that automount attempts a sendto() call, waits
for it to return (about 5 seconds), and then trys it again, usually
about 10-12 times total.

Is there any way to prevent automountd from reattempting the
connection this many times? Once or twice would be sufficient!

I've looked at the timeo and retrans statistics in nfsstat -- they
each are incremented by 1 between the start of the automount attempt
and the 1 minute delay. So, it appears that the NFS client thinks it's
only trying once (and failing).

I am running HPUX 10.20 with the latest patches. As you can probably
tell from my message, I am just shooting around in the dark. Any help
would be appreciated.

Thanks,
Matt Banek
L3 Communications IS
1 REPLY 1
James A. Donovan
Honored Contributor

Re: HPUX 10.20 - long delays when attempting to access file on unavailable automounted file system

I don't think you can change this behavior. But if you can affect it at all, I would think it would be by tweaking your tcp/ip parameters using ndd.

You can get much more detailed information on what is happening by adding -Tvvv to the list of automountd parameters. You'll see a lot of extra info in your /var/adm/automount.log and your syslog file.

You could also try using tusc to profile the automountd daemon to see the system calls being made.
Remember, wherever you go, there you are...