11.23 NFS and AutoFS

[Pete Randall]

My new 11.23 environment seems to be having some NFS difficulties. It's behaving just fine with other HP-UX environments with automounts happening in both directions (from the 11.23 box to the others and from the others to the 11.23 box).

However, our Novell file server at the heart of our LAN doesn't seem to want to play nicely. It appears that the 11.23 box attempts to automount as it should but we can't read anything past the exported LAN directory:

# cd /nfs/farm/prod1
# ll
total 0
dr-xr-xr-x 1 root root 1 Feb 13 13:13 BTID
dr-xr-xr-x 1 root root 1 Feb 13 13:13 WP
dr-xr-xr-x 1 root root 1 Feb 13 13:13 allflex
dr-xr-xr-x 1 root root 1 Feb 13 13:13 common
dr-xr-xr-x 1 root root 1 Feb 13 13:13 dv
dr-xr-xr-x 1 root root 1 Feb 13 13:13 m9
dr-xr-xr-x 1 root root 1 Feb 13 13:13 ped
dr-xr-xr-x 1 root root 1 Feb 13 13:13 shared
dr-xr-xr-x 1 root root 1 Feb 13 13:13 tristar
dr-xr-xr-x 1 root root 1 Feb 13 13:13 web
# mount |grep common
/nfs/farm/prod1/common on -hosts ignore,direct,nosuid,hard,intr,bg,nest,dev=9 on
Fri Feb 13 13:13:07 2009
# ll common
common unreadable
total 0


On my other 11.11 machines, this works just fine. What am I missing?

Another question: how do I check which version of NFS I have?


Pete

[Victor Fridyev]

What about nfsstat ?

[Pete Randall]

Sorry, Victor, I'm not following. What do you mean, "what about nfsstat"?


Pete

[Victor Fridyev]

Pete,

It seems, nfsstat commanf gives NFS version.

[Pete Randall]

Well, yes, it does seem to indicate that I have either Version 2 or Version 3.


Pete

[Pete Randall]

OK, I see - it's telling me it will support both v2 and v3. Presumably that's as a client and as a server, I would expect it to be v3?


Pete

[Victor Fridyev]

V3 works for sure.
BTW, 11.23 has some problems with autofs, e.g. I have not succeeded to change AUTOMASTER name, i.e. I edited /etc/auto.master, but the comp read /etc/auto_master, even the variable was changed in the configuration

Good Luck

[Pete Randall]

Ping!

[Pete Randall]

Maybe more info would be helpful!

Here's the AutoFS trace entries of the mount attempt:

Feb 16 12:32:50 t134 LOOKUP REQUEST: Mon Feb 16 12:32:50 2009
Feb 16 12:32:50 t134 name=farm[] map=-hosts opts=nosuid,hard,intr,bg,vers=2 path=/nfs direct=0
Feb 16 12:32:50 t134 LOOKUP REPLY : status=0
Feb 16 12:32:50 t1 MOUNT REQUEST: Mon Feb 16 12:32:50 2009
Feb 16 12:32:50 t1 name=farm[] map=-hosts opts=nosuid,hard,intr,bg,vers=2 path=/nfs direct=0
Feb 16 12:32:50 t1 MOUNT REPLY : status=0, AUTOFS_ACTION
Feb 16 12:32:50 t135 MOUNT REQUEST: Mon Feb 16 12:32:50 2009
Feb 16 12:32:50 t135 name=farm[/prod1] map=-hosts opts=nosuid,hard,intr,bg,vers=2 path=/nfs/farm
/prod1 direct=1
Feb 16 12:32:50 t135 MOUNT REPLY : status=0, AUTOFS_ACTION
Feb 16 12:32:50 t1 MOUNT REQUEST: Mon Feb 16 12:32:50 2009
Feb 16 12:32:50 t1 name=farm[/prod1/common] map=-hosts opts=nosuid,hard,intr,bg,vers=2 path=/n
fs/farm/prod1/common direct=1
Feb 16 12:32:50 t1 Port numbers are 666, 666
Feb 16 12:32:50 t1 Port match
Feb 16 12:32:50 t1 MOUNT REPLY : status=13, AUTOFS_DONE


And the exported directories, as seen from the NFS client:

# showmount -e farm
export list for farm:
/image1/vol1 (everyone)
/prod1/BTID/download (everyone)
/prod1/common (everyone)


So, why is "access denied"?


Pete

[T G Manikandan]

Pete,

check this

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1045794

[Pete Randall]

Sorry but I don't see where that applies to my situation.


Pete

[Steven E. Protter]

Shalom,

11.11 works with this Novell server.
11.23 does not.

I would check the patch database against the error message and the keyword Novell. Read the notes see if there is something that applies.

If in a hurry, put together a large nfs patch set, install it and hope the shotgun gets the problem.

I used to work with victor and if you published the configuration files he or I might be able to spot something.

SEP

[Pete Randall]

No patch hits with any keywords other than just "nfs", so I applied all the latest nfs and autofs patches. Still no joy.

;^(


Pete

[Dennis Handly]

Have you bypassed the automounter and tried hard mounting common? Do it somewhere else to check, /var/tmp/common/.

[Dave Olker]

Hi Pete,

Let's take a look at the working and failing system to see what they're doing.

On the 11.11 NFS client, issue the commands:

# nfsstat -m
# rpcinfo -p

I'd also like to see the AutoFS maps used to mount the filesystem from the Novell server.

Then on the 11.23 NFS client, issue the command:

# nfsstat -m

and collect the AutoFS maps used by the 11.23 system to mount the Novell filesystem (unless they are the same maps being shared via NIS or something).

Regards,

Dave

[Pete Randall]

On the 11.11 client:

# nfsstat -m
# rpcinfo -p 130.1.0.245
program vers proto port service
100000 2 udp 111 rpcbind
100000 2 tcp 111 rpcbind
100007 2 udp 902 ypbind
100009 1 udp 903 yppasswdd
100009 1 tcp 846 yppasswdd
100004 2 udp 904 ypserv
100004 1 udp 961 ypserv
100004 1 tcp 731 ypserv
100005 1 udp 1234 mountd
100005 2 udp 1234 mountd
100005 3 udp 1234 mountd
100005 1 tcp 1234 mountd
100005 2 tcp 1234 mountd
100005 3 tcp 1234 mountd
100003 2 udp 2049 nfs
100003 2 tcp 2049 nfs
100003 3 udp 2049 nfs
100003 3 tcp 2049 nfs
100021 1 udp 32779 nlockmgr
100021 2 udp 32779 nlockmgr
100021 3 udp 32779 nlockmgr
100021 4 udp 32779 nlockmgr
100021 1 tcp 32779 nlockmgr
100021 2 tcp 32779 nlockmgr
100021 3 tcp 32779 nlockmgr
100021 4 tcp 32779 nlockmgr
100024 1 udp 32778 status
100024 1 tcp 32778 status
# cat /etc/auto_master
/nfs -hosts -nosuid,hard,intr,bg


On the 11.23 client:

# nfsstat -m
# cat /etc/auto_master
/nfs -hosts -nosuid,hard,intr

Both AutoFS maps used to have "bg" specified until an hour or so ago when I read that it wasn't allowed for an automounted directory. Taking it out didn't help.


Pete

[Dave Olker]

I guess I should have made it clear that before issuing the "nfsstat -m" commands you'd need to actually mount the filesystems so they'd show up in the output. :)

So, on both systems please force AutoFS to mount the filesystem first then issue the nfsstat -m command on both systems.

Also, can you tell me if the 11.11 system is using the original automounter, the original (ONC 1.2) AutoFS, or the Enhanced AutoFS?

Dave

[Pete Randall]

The 11.11 system is using the original automounter and I attempted to mount before issuing the nfsstat command.

11.11:
# ll /nfs/farm/prod1
total 0
dr-xr-xr-x 3 root root 96 Feb 17 13:36 BTID
dr-xr-xr-x 4 root root 96 Feb 17 13:36 WP
drwxrwxrwx 1 root root 4096 Feb 5 2007 allflex
drwxrwxrwx 1 root root 4096 Feb 13 13:12 common
drwxrwxrwx 1 root root 4096 Feb 16 12:54 dv
drwxrwxrwx 1 root root 4096 Dec 13 2005 m9
drwxrwxrwx 1 root root 4096 Jun 6 2008 ped
dr-xr-xr-x 4 root root 96 Feb 17 13:36 shared
drwxrwxrwx 1 root root 4096 Feb 1 2008 tristar
drwxrwxrwx 1 root root 4096 Mar 1 2004 web
# nfsstat -m
#

11.23
# ll /nfs/farm/prod1
total 0
dr-xr-xr-x 1 root root 1 Feb 17 13:30 BTID
dr-xr-xr-x 1 root root 1 Feb 17 13:30 WP
dr-xr-xr-x 1 root root 1 Feb 17 13:30 allflex
dr-xr-xr-x 1 root root 1 Feb 17 13:30 common
dr-xr-xr-x 1 root root 1 Feb 17 13:30 dv
dr-xr-xr-x 1 root root 1 Feb 17 13:30 m9
dr-xr-xr-x 1 root root 1 Feb 17 13:30 ped
dr-xr-xr-x 1 root root 1 Feb 17 13:30 shared
dr-xr-xr-x 1 root root 1 Feb 17 13:30 tristar
dr-xr-xr-x 1 root root 1 Feb 17 13:30 web
# nfsstat -m
#


Pete

[Dave Olker]

On the working 11.11 system, try this:

# cd /nfs/farm/prod1/common
# nfsstat -m

Hopefully that will show the NFS mount.

Looking at the AutoFS log you provided:

AUTOFS_ACTION
Feb 16 12:32:50 t1 MOUNT REQUEST: Mon Feb 16 12:32:50 2009
Feb 16 12:32:50 t1 name=farm[/prod1/common] map=-hosts opts=nosuid,hard,intr,bg,vers=2 path=/n
fs/farm/prod1/common direct=1
Feb 16 12:32:50 t1 Port numbers are 666, 666
Feb 16 12:32:50 t1 Port match
Feb 16 12:32:50 t1 MOUNT REPLY : status=13, AUTOFS_DONE

The NFS server is denying the last mount request with "status=13". That too would explain why the overall NFS mount is failing.

As Dennis suggested earlier, try doing a manual NFS mount on the 11.23 system just to take AutoFS out of the picture for the moment:

11.23> mkdir /test1
11.23> mount farm:/prod1/common /test1

If that works, issue the "nfsstat -m" command so I can see what got mounted.

Dave

[Pete Randall]

Dennis & Dave,

11.11# cd /nfs/farm/prod1/common
11.11# nfsstat -m
11.11#

11.23# mkdir /test1
11.23# mount farm:/prod1/common /test1
No such file or directory


The only time I can get anything out of the nfsstat command is on the 11.23 system when I cause one of the 11.11 systems to get automounted.


Pete

[Dennis Handly]

>11.23# mkdir /test1
>11.23# mount farm:/prod1/common /test1
>No such file or directory

Hmm, I don't think I've seen that error before with mount?

Can you try those commands on your 11.11 system and do Dave's:
nfsstat -m

[Pete Randall]

# uname -r
B.11.11
# mkdir /test1
# mount farm:/prod1/common /test1
# nfsstat -m
/test1 from farm:/prod1/common (Addr 130.1.0.245)
Flags: vers=3,proto=tcp,auth=unix,hard,intr,link,symlink,devs,rsize=32768,wsi
ze=32768,retrans=5
All: srtt= 0 ( 0ms), dev= 0 ( 0ms), cur= 0 ( 0ms)

#


Pete

[Dave Olker]

Pete,

Your last test proves the 11.11 client can mount the filesystem without issue and the 11.23 client cannot. Looking back at the AutoFS log file data you posted, the Novell server is rejecting the MOUNT request from the 11.23 client with a status of 13.

I think you need to focus on the Novell server and determine why it is rejecting the mount request. Can you print out exactly how the filesystem is exported on the Novell box? I'm not familiar with Novell so I don't know the proper commands to find this. Is there an equivalent of /etc/xtab on Novell?

My guess is the Novell box has some kind of restriction on which NFS clients can mount the filesystem and the 11.23 client is not in the approved access list.

Regards,

Dave

[Pete Randall]

Dave,

That was my conclusion from the very beginning but the Network Admin in charge of the "farm" insists that nothing is restricted, as evidenced by the showmount output. I'm not intimately familiar with Netware either but wouldn't you agree that showmount certainly looks like it's unrestricted?


Pete

[Dave Olker]

Pete,

I've been burned many, many times by trusting what showmount reports. I never rely on that information. My preference would be to get on the NFS server and look at exactly how the filesystem is currently shared.

Also, even if the filesystem is shared properly, the server still seems to be the one rejecting the MOUNT request. If it's not due to access rights then there's some other reason. Perhaps the Novell server cannot resolve the hostname/IP address of the 11.23 system and it fails any client for that reason.

Bottom line, there could be any number of reasons why the server is rejecting the 11.23 client. If this were an HP-UX NFS server the answer would be very simple - turn on debug rpc.mountd logging and collect the log file from the failure. That would tell me exactly why the server rejected a mount request. Is there such a detailed logging facility on the Novell system?

If you don't have access to the Novell system, one thing you could collect would be a nettl trace on the 11.23 system of the failed mount request. I would also collect a nettl trace on the 11.11 system of the working mount request. If you want to collect those traces and send me the raw trace files I can at least look at the MOUNT request from both systems and verify that there is nothing inherently different about the MOUNT requests themselves and then it will really be up to the Novell folks to figure out why the server is rejecting one and accepting the other.

Dave