nfs mount stopped working (hp ux 11.23)

[Donald Thaler]

we've been using an nfs mount for the past 3-4 weeks and all of a sudden i get "nfs server not responding" when trying to access the nfs mount... checked the server and the nfs process is running.. i ran the /sbin/init.d/nfs.server start command and it basically told me everything was already running. the only thing that has changed that i'm aware of is that we started up the snmp process ... what else can i look at

[Dennis Handly]

Have you looked at the NFS server? Can you ping it?

[Donald Thaler]

yes i can ping it..i can do an nslookup on the nfs server from other servers

[Chandrahasa s · Accepted solution]

HI,

Start the nfs services on server as like

/sbin/init.d/nfs.client stop
/sbin/init.d/nfs.server stop
/sbin/init.d/nfs.core stop


/sbin/init.d/nfs.core start
/sbin/init.d/nfs.server start
/sbin/init.d/nfs.client start

check showmount -e shows any exporred mount points

Chandra

[Modris Bremze]

Check if all the necessary services really are started with rpcinfo on both nodes.

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1334452

[Donald Thaler]

stopping and starting the nfs.client/server/core fixed the problem...

the next question is what could have happened that caused it to stop ??

these are the events as best i can remember that took place between the time nfs was working and when it stopped:

i restarted snmp, i updated some kernel entries, but had not rebooted the server..(yet a who -r showed a reboot about the time i did the kernel changes ?)

[Bill Hassell]

>> the next question is what could have happened that caused it to stop ??

Many sysadmins ask the question: What does it take to keep NFS working?

A great book on NFS for HP-UX is Optimizing NFS Performance: Tuning and Troubleshooting NFS on HP-UX Systems by Dave Olker. NFS is critically dependent on the server and client plus every step between the two. There's not a lot you can determine after a disconnect, and only a few things you can turn on before the next failure. Ideally, there would be a nettl or Wireshark trace of the network so you can see some of the possible problems. Even then, a flaky router may have caused the problem. NFS doesn't seem to recover reliably from network problems. Try to connect directly to bypass switches and routers.

The good news (as you've seen) is that restarting NFS services will get things repaired. And of course, the very latest patches for networking and NFS are mandatory.

[Donald Thaler]

i've been reading other threads on nfs problems and in some cases a reboot has actually been halted by an nfs problem and the soltuion that i've seen has been to press the ctl key and the pipe key or the ctl key and the 'c' key that kills the hung startup script. since i use a cde interface, exceed, to communicate with the hp unix server how would i know if the startup process was hung ? is there some way i should do the restart that i would know if the startup process was hung ??

am i tilting a windmills ??

[Dennis Handly]

>since I use a CDE interface,

You don't use CDE when booting. It is either the ITE or some type of xterm/serial console window.

>is there some way I should do the restart that I would know if the startup process was hung?

Have you logged on to the MP in console mode so you can see the progress?

[Donald Thaler]

actually i do use cde... i do a shutdown -r, and then i wait for the server name to appear in the exceed management window and then i know its back up...i assume the management port would be a better way to shutdown and restart the server ...

is there anyway of displaying what the actual mount command looked like when it was created ? i created an nfs mount using a mount -F nsf -o hard..... but i can't rember the full construction of the command or where it might be located ??

[Dennis Handly]

>actually I do use CDE.

But not on the console.

>I assume the management port would be a better way to shutdown and restart the server

Yes, much.

>is there anyway of displaying what the actual mount command looked like when it was created?

Well, if it is mounted on boot, then you can look at /etc/fstab.
Or you can look at /etc/mnttab. Or must type mount to format /etc/mnttab.

>I created an nfs mount using a mount -F nfs -o hard... but I can't remember the full construction of the command or where it might be located?

If you did this manually on a previous boot, this is lost, unless it is around in your shell history file.

[Donald Thaler]

somehow i lost the nfs mount again, but this time doing the stop and start on the client,server and core didn't help.. when i do a showmount -e i see:
/u06/ldif (everyone)
/u06/rman_backup (everyone)

a showmount -a:
(anon):/u06/ldif
(anon):/u06/rman_backup
sc0gllb1.lccc.edu:/u06/TEST_NSF


but when i try to access them from my other server it tells me the nfs server is not responding...

i've done an exportfs -a...

i modified some kernal entries, but haven't rebooted yet... would this somehow affect the nfs mount ??

is there something else i can look at that would give me a clue as to what's going on

[Donald Thaler]

i did an nfsstat -m and it returned the specs that i used to create the nfs mount:
nfsstat -m
/u03/rman_restore from apollo:/u06/rman_backup (Addr 10.2.2.20)
Flags: vers=3,proto=tcp,auth=unix,hard,printed,intr,noac,link,symlink,devs,rsize=32768,wsize=32768,retrans=5

[TwoProc]

Donald, you can get that error when an NFS mount point gets extremely busy. You may not have enough bandwidth to do what it is you're doing there. There are also proper ways to get mounts working that minimize the amount of traffic needed to keep that connection sync'd up. Please do use and read the book that Bill Hassell told you about, I have it and it does make it much easier to understand what it is you're trying do, and how to do it better.

How does your networking look?

[Donald Thaler]

the mount points are not busy... they are used sporadically in a strictly test environment... as far as How does your networking look? i'm not sure what you'e looking for.. all the servers are on a local domain.. the nfs mount was created on our production server, and our backup server attempts to read files from the nfs mount... this used to work, and then it stopped working..?? i'll look for that book... i just thought being a novice at nfs maybe there was some simple process i missed in addition to stopping and starting the nfs client,core,server..

[Donald Thaler]

if you stop the nfs server, then kill
sfsktcpd, then restart the nfs server the nfs mount comes back ???

[TwoProc]

So, this production server interface that you're connecting to - it's not busy? Even if your client connection isn't busy itself, the physical port on production that is hosting your nfs server may be, and if it gets overloaded at peaks of the day, then you'd see what you're seeing.