I have two Solaris 8 boxes on different networks using NIS+. There are
no other suns in the network. One of them (call it aserver) is the
homedir server. The other (call it aclient) mounts the homedirs via
autofs and the auto_home NIS+ table. Until yesterday everything was
working OK. Then suddenly I couldn't see the home directories from aclient.

After some attempts to troubleshoot including restarts of RPC, NFS and
autofs on both machines I see that I cannot automount anything served by
aserver on the client aclient. Normal NFS mounts seem to work.

aclient# ls /home/user
/home/user: Permission denied

In the logs on aclient I get lots of messages as this:

aclient automountd[5272]: [ID 784820 daemon.error] server aserver not
responding

I have tried everything from the NFS troubleshooting guides. Everything
works OK except this

aclient# rpcinfo -T udp aserver nfs
rpcinfo: RPC: Timed out
program 100003 version 0 is not available

aserver# rpcinfo -T udp aserver nfs
program 100003 version 2 ready and waiting
program 100003 version 3 ready and waiting

aclient#rpcinfo -T tcp aserver nfs
program 100003 version 2 ready and waiting
program 100003 version 3 ready and waiting

aserver#rpcinfo -T tcp aserver nfs
program 100003 version 2 ready and waiting
program 100003 version 3 ready and waiting

(Just for the record, the reverse gives exactly the same:

aserver# rpcinfo -T udp aclient nfs
rpcinfo: RPC: Timed out
program 100003 version 0 is not available

and so on...)

So for NFS udp from aclient to aserver (and back) is not working, but
tcp works. As for the other services (mountd for example) -- I get no
problems with udp and tcp.

As I try to trace it with snoop (on both machines), I see this on the
client and the server:

client#rpcinfo -T server nfs

client -> server PORTMAP C GETPORT prog=100003 (NFS) vers=0 proto=UDP
server -> client PORTMAP R GETPORT port=2049

But:

client#rpcinfo -T server nfs 3

snoop on the client:

client -> server PORTMAP C GETPORT prog=100003 (NFS) vers=0 proto=UDP
server -> client PORTMAP R GETPORT port=2049
client -> server NFS C NULL3

snoop on the server:
client -> server PORTMAP C GETPORT prog=100003 (NFS) vers=0 proto=UDP
server -> client PORTMAP R GETPORT port=2049

The third packet does not seem to come. It looks like a network problem
but I cannot imagine of what kind. And why only with NFS/autofs?!

Any help will be highly appreciated.

Regards,
E.P.

In article <2tkck2F20nm89U1@uni-berlin.de>, emil.petkov@usa.net says...
> I have two Solaris 8 boxes on different networks using NIS+. There are
> no other suns in the network. One of them (call it aserver) is the
> homedir server. The other (call it aclient) mounts the homedirs via
> autofs and the auto_home NIS+ table. Until yesterday everything was
> working OK. Then suddenly I couldn't see the home directories from aclient.

So what has changed? In cases like this it is almost always because someone
made a change. Did someone change a router setting or put a firewall in
place?
--
Mike Miller
If all else fails - READ THE INSTRUCTIONS!
or if you like
"If all else fails - THROW HARDER" Robert Smith(pro bowler)

Mike Miller wrote:

> In article <2tkck2F20nm89U1@uni-berlin.de>, emil.petkov@usa.net says...
>
>>I have two Solaris 8 boxes on different networks using NIS+. There are
>>no other suns in the network. One of them (call it aserver) is the
>>homedir server. The other (call it aclient) mounts the homedirs via
>>autofs and the auto_home NIS+ table. Until yesterday everything was
>>working OK. Then suddenly I couldn't see the home directories from aclient.
>
>
> So what has changed? In cases like this it is almost always because someone
> made a change. Did someone change a router setting or put a firewall in
> place?

This is a good question. I have asked the IT services about that but my
experience shows that it will take a while before they answer. However,
this is what I know:

There is only one router between the two networks. If I scan the server
from the client with nmap I see that the nfs udp port (2049/udp) as well
as rpc are open.

aclient# rpcinfo -p aserver

lists all necessary programs.

E.

In article <2tkj0pF211sr0U1@uni-berlin.de>, emil.petkov@usa.net says...
> Mike Miller wrote:
>
> > In article <2tkck2F20nm89U1@uni-berlin.de>, emil.petkov@usa.net says...
> >
> >>I have two Solaris 8 boxes on different networks using NIS+. There are
> >>no other suns in the network. One of them (call it aserver) is the
> >>homedir server. The other (call it aclient) mounts the homedirs via
> >>autofs and the auto_home NIS+ table. Until yesterday everything was
> >>working OK. Then suddenly I couldn't see the home directories from aclient.
> >
> >
> > So what has changed? In cases like this it is almost always because someone
> > made a change. Did someone change a router setting or put a firewall in
> > place?
>
> This is a good question. I have asked the IT services about that but my
> experience shows that it will take a while before they answer. However,
> this is what I know:
>
> There is only one router between the two networks. If I scan the server
> from the client with nmap I see that the nfs udp port (2049/udp) as well
> as rpc are open.
>
> aclient# rpcinfo -p aserver
>
> lists all necessary programs.
>
> E.
>
In your first post you said normal NFS mounts worked. Have you tried
mounting one of the home directories by hand? Also, are you using netgroups
anywhere? And how about the underlying permissions on the mount points?
Since it doesn't look like your network infrastructure changed these are
some items that can cause permission denied on automounts.
--
Mike Miller
If all else fails - READ THE INSTRUCTIONS!
or if you like
"If all else fails - THROW HARDER" Robert Smith(pro bowler)

Mike Miller wrote:
> In article <2tkj0pF211sr0U1@uni-berlin.de>, emil.petkov@usa.net says...
>
>>Mike Miller wrote:
>>
>>
>>>In article <2tkck2F20nm89U1@uni-berlin.de>, emil.petkov@usa.net says...
>>>
>>>
>>>>I have two Solaris 8 boxes on different networks using NIS+. There are
>>>>no other suns in the network. One of them (call it aserver) is the
>>>>homedir server. The other (call it aclient) mounts the homedirs via
>>>>autofs and the auto_home NIS+ table. Until yesterday everything was
>>>>working OK. Then suddenly I couldn't see the home directories from aclient.
>>>
>>>
>>>So what has changed? In cases like this it is almost always because someone
>>>made a change. Did someone change a router setting or put a firewall in
>>>place?
>>
>>This is a good question. I have asked the IT services about that but my
>>experience shows that it will take a while before they answer. However,
>>this is what I know:
>>
>>There is only one router between the two networks. If I scan the server
>>from the client with nmap I see that the nfs udp port (2049/udp) as well
>>as rpc are open.
>>
>>aclient# rpcinfo -p aserver
>>
>>lists all necessary programs.
>>
>>E.
>>
>
> In your first post you said normal NFS mounts worked. Have you tried
> mounting one of the home directories by hand? Also, are you using netgroups
> anywhere? And how about the underlying permissions on the mount points?
> Since it doesn't look like your network infrastructure changed these are
> some items that can cause permission denied on automounts.

Yes, I tried. I cannot umount /home (see below), but I mounted a homedir
under some other mount point and it is perfectly readable. Yes, I use
netgroups but I tried without them as well (exported homes for everyone
instead of a particular netgroup) -- the situation is the same. The
mount point /home seems ok -- root:root 555.

I stopped the autofs but still cannot get hold of /home. 'umount -f
/home' hangs.

It could be the network infrastructure -- the IT service dept asked me
where are the two suns connected and they were going to look at it.
However, it seems really strange to me that only NFS UDP gives me
problems. But perhaps I am missing something.

E.P.

In article <2tmp5sF22m289U1@uni-berlin.de>, emil.petkov@usa.net says...
> Mike Miller wrote:
> > In article <2tkj0pF211sr0U1@uni-berlin.de>, emil.petkov@usa.net says...
> >
Do you use any other automounts? Maybe you could try (for a test) putting
the homedir mounts into another automount map. It won't fix the problem but
it could help track it down. If it doesn't work it points more to
infrastructure and if it does work then there is some issue with auto_home.
Another thought that comes to mind is your overall NIS+ health. I assumed
that NIS+ was OK because of the small environment.
--
Mike Miller
If all else fails - READ THE INSTRUCTIONS!
or if you like
"If all else fails - THROW HARDER" Robert Smith(pro bowler)

Emil Petkov <emil.petkov@usa.net> wrote in
news:2tmp5sF22m289U1@uni-berlin.de:

..........
for everyone instead of a particular netgroup) -- the situation is the
> same. The mount point /home seems ok -- root:root 555.
>
...........


regardless of any other points - the mount point should have at least 755
instead of 555. i did experience problems in the past having less than 755.

robert