opensolaris: cant mount nfs disk - Solaris

Hi,

The other day I installed OpenSolaris but so far I'm unable to mount my
nfs disk:

> # mount -F nfs Kaapstad:/admin /mnt/Kaapstadadmin/
> nfs mount: Kaapstad:/admin: No such file or directory

Kaapstad is in /etc/hosts, Kaapstad:/admin is in /etc/vfstab. Other
machines have no problem mounting the disk, and this machine is in the
exports-list.
Any clue?

Thanks.

Huub wrote:
> Hi,
>
> The other day I installed OpenSolaris but so far I'm unable to mount my
> nfs disk:
>
>> # mount -F nfs Kaapstad:/admin /mnt/Kaapstadadmin/
>> nfs mount: Kaapstad:/admin: No such file or directory
>
> Kaapstad is in /etc/hosts, Kaapstad:/admin is in /etc/vfstab. Other
> machines have no problem mounting the disk, and this machine is in the
> exports-list.
> Any clue?
>
See if the automounter can see it, does anything show up under
/net/Kaapstad?

--
Ian Collins.

On Sat, 10 May 2008 06:40:45 +0200, Huub wrote:

>> # mount -F nfs Kaapstad:/admin /mnt/Kaapstadadmin/
>> nfs mount: Kaapstad:/admin: No such file or directory

Does your NFS server support NFS-v4?

Edit /etc/default/nfs

NFS_CLIENT_VERSMAX=3

> Does your NFS server support NFS-v4?
>
> Edit /etc/default/nfs
>
> NFS_CLIENT_VERSMAX=3
>

Didn't think about this...this did it! Thank you.

On 2008-05-11 08:36:34 +0100, Huub said:

>> Does your NFS server support NFS-v4?
>>
>> Edit /etc/default/nfs
>>
>> NFS_CLIENT_VERSMAX=3
>>
>
> Didn't think about this...this did it! Thank you.

Is the Solaris NFSv4 client doing something wrong, or the NFSv4 server
doing something wrong? It is a pity to have to downgrade to v3...

Cheers,

Chris

On Sun, 11 May 2008 10:44:26 +0100, Chris Ridd wrote:

> Is the Solaris NFSv4 client doing something wrong, or the NFSv4 server
> doing something wrong? It is a pity to have to downgrade to v3...

The Solaris client *should* adapt to the capability of the server without
a kluge like that. It does not.

On 2008-05-11 13:32:32 +0100, Dave Uhring said:

> On Sun, 11 May 2008 10:44:26 +0100, Chris Ridd wrote:
>
>> Is the Solaris NFSv4 client doing something wrong, or the NFSv4 server
>> doing something wrong? It is a pity to have to downgrade to v3...
>
> The Solaris client *should* adapt to the capability of the server without
> a kluge like that. It does not.

I suppose there are a couple of things to ask then: a) why isn't NFSv4
working, and b) as it isn't, why isn't the downgrade to NFSv3 working?

Cheers,

Chris

On Sun, 11 May 2008 13:43:49 +0100, Chris Ridd wrote:
> On 2008-05-11 13:32:32 +0100, Dave Uhring said:
>> On Sun, 11 May 2008 10:44:26 +0100, Chris Ridd wrote:
>>
>>> Is the Solaris NFSv4 client doing something wrong, or the NFSv4 server
>>> doing something wrong? It is a pity to have to downgrade to v3...
>>
>> The Solaris client *should* adapt to the capability of the server without
>> a kluge like that. It does not.
>
> I suppose there are a couple of things to ask then: a) why isn't NFSv4
> working, and b) as it isn't, why isn't the downgrade to NFSv3 working?

I don't know how an NFSv4-only client can connect to an NFSv3 server
unless the client is smart enough to use NFSv3. The one in Solaris is not
that smart. Neither S10u4, S10u5, OpenSolaris nor any of the Nevada
builds are capable of mounting an NFSv3 share without modifying
/etc/default/nfs.

Hack /etc/default/nfs and the Solaris client easily mounts the NFSv3 share.

On 2008-05-11 16:16:59 +0100, Dave Uhring said:

> On Sun, 11 May 2008 13:43:49 +0100, Chris Ridd wrote:
>> On 2008-05-11 13:32:32 +0100, Dave Uhring said:
>>> On Sun, 11 May 2008 10:44:26 +0100, Chris Ridd wrote:
>>>
>>>> Is the Solaris NFSv4 client doing something wrong, or the NFSv4 server
>>>> doing something wrong? It is a pity to have to downgrade to v3...
>>>
>>> The Solaris client *should* adapt to the capability of the server without
>>> a kluge like that. It does not.
>>
>> I suppose there are a couple of things to ask then: a) why isn't NFSv4
>> working, and b) as it isn't, why isn't the downgrade to NFSv3 working?
>
> I don't know how an NFSv4-only client can connect to an NFSv3 server

OK, I didn't think it was an NFSv4-only client.

> unless the client is smart enough to use NFSv3. The one in Solaris is not
> that smart. Neither S10u4, S10u5, OpenSolaris nor any of the Nevada
> builds are capable of mounting an NFSv3 share without modifying
> /etc/default/nfs.

They can't even mount an NFSv4 share from some kinds of servers :-(

> Hack /etc/default/nfs and the Solaris client easily mounts the NFSv3 share.

Yep, that's pretty much the first thing we do with a new box...

Cheers,

Chris

On Sun, 11 May 2008 16:39:15 +0000, Andrew Gabriel wrote:

> Solaris will fall back to NFSv3 and NFSv2 if NFSv4 isn't supported.
> If the server does have NFSv4 enabled but it's not working right, then
> there's no fallback -- that's not what the fallback is intended for.

OK, perhaps the Linux NFS server is the real problem then. I thought that
I had disabled NFSv4 on that server but

[root@newton /proc/17856]# cat status
Name: nfsd4

I never did trust Linux NFS very much but that machine is my print server
also and the printer requires HPLIP, which will not build on Solaris.

On 2008-05-11 18:14:22 +0100, Dave Uhring said:

> On Sun, 11 May 2008 16:39:15 +0000, Andrew Gabriel wrote:
>
>> Solaris will fall back to NFSv3 and NFSv2 if NFSv4 isn't supported.
>> If the server does have NFSv4 enabled but it's not working right, then
>> there's no fallback -- that's not what the fallback is intended for.
>
> OK, perhaps the Linux NFS server is the real problem then. I thought that
> I had disabled NFSv4 on that server but
>
> [root@newton /proc/17856]# cat status
> Name: nfsd4
>
> I never did trust Linux NFS very much but that machine is my print server
> also and the printer requires HPLIP, which will not build on Solaris.

Do we know what the exact NFS incompatibility is here?

Cheers,

Chris

On Sun, 11 May 2008 17:33:38 +0000, Andrew Gabriel wrote:

> You might want to glance at this in case it's any help...
> http://blogs.sun.com/macrbg/entry/so...ent_mount_from

I added the fsid=0 argument to /etc/exports on the Linux box, restarted
the nfs-kernel-server. /etc/exports on that box:

/srv (fsid=0,sync,insecure,no_root_squash,rw,nohide)

Edited /etc/default/nfs on my nvb_87 system to restore default client
behavior and:

[root@einstein /etc/default]$ mount newton:/srv /mnt
nfs mount: newton:/srv: No such file or directory

Restore /etc/default/nfs to NFS_CLIENT_VERSMAX=3 and:

[root@einstein /etc/default]$ mount newton:/srv /mnt
[root@einstein /etc/default]$

At least NFSv3 is working correctly on the Linux server which is one HELL
of an improvement over years past! But I have set up another machine
running Open Solaris 200805 using ZFS and when I get a little more
confidence in it I'll just stop the NFS server on that Linux box.

Dave Uhring wrote:
> On Sun, 11 May 2008 16:39:15 +0000, Andrew Gabriel wrote:
>
>> Solaris will fall back to NFSv3 and NFSv2 if NFSv4 isn't supported.
>> If the server does have NFSv4 enabled but it's not working right, then
>> there's no fallback -- that's not what the fallback is intended for.
>
> OK, perhaps the Linux NFS server is the real problem then. I thought that
> I had disabled NFSv4 on that server but
>
> [root@newton /proc/17856]# cat status
> Name: nfsd4
>
> I never did trust Linux NFS very much but that machine is my print server
> also and the printer requires HPLIP, which will not build on Solaris.
>
The Linux NFSv4 server does *not* fall back to NFSv3 when it is not set
up to do authentication and all the trimmings of NFSv4. I can't
remember the exact sequence, but it basically sends back a failure
status rather than a "try again".

IOW - out of the box *no* NFSv4 client can connect to it, /not even it's
own/!

Cheers,
Gary B-)

--
__________________________________________________ ____________________________
Armful of chairs: Something some people would not know
whether you were up them with or not
- Barry Humphries

On Mon, 12 May 2008 09:05:46 +1000, Gary R. Schmidt wrote:

> The Linux NFSv4 server does *not* fall back to NFSv3 when it is not set
> up to do authentication and all the trimmings of NFSv4. I can't
> remember the exact sequence, but it basically sends back a failure
> status rather than a "try again".

Very good point and one which I had not considered.

> IOW - out of the box *no* NFSv4 client can connect to it, /not even it's
> own/!

However, the Solaris client does connect with the Solaris server using
NFSv3 when the prerequisites for NFSv4 have not been set on the server
and *without* the hack on /etc/default/nfs.

Andrew Gabriel wrote:
> In article ,
> "Gary R. Schmidt" writes:
>> Dave Uhring wrote:
>>> On Sun, 11 May 2008 16:39:15 +0000, Andrew Gabriel wrote:
>>>
>>>> Solaris will fall back to NFSv3 and NFSv2 if NFSv4 isn't supported.
>>>> If the server does have NFSv4 enabled but it's not working right, then
>>>> there's no fallback -- that's not what the fallback is intended for.
>>> OK, perhaps the Linux NFS server is the real problem then. I thought that
>>> I had disabled NFSv4 on that server but
>>>
>>> [root@newton /proc/17856]# cat status
>>> Name: nfsd4
>>>
>>> I never did trust Linux NFS very much but that machine is my print server
>>> also and the printer requires HPLIP, which will not build on Solaris.
>>>
>> The Linux NFSv4 server does *not* fall back to NFSv3 when it is not set
>> up to do authentication and all the trimmings of NFSv4. I can't
>> remember the exact sequence, but it basically sends back a failure
>> status rather than a "try again".
>
> That's correct behaviour -- NFS servers have no way to fall back.
> It's the clients which fall back, but this is to cope with systems
> which haven't implemented NFSv4 (or NFSv3), not for working around
> buggy implementations. A client has a not unreasonable expecation
> that if a server offers NFSv4, it's a working implementation.
>
Yes, bad explanation by me, I'll try again...

The Linux NFSv4 server, when receiving a mount request from *any* NFSv4
client before it has had its authentication configured returns to the
client "no such share," or something to that effect which the client
sees as "don't try again, it won't work," instead of sending "I don't
like you" or "I don't understand you," which would allow the client to
try again at NFSv3 (or v2) and succeed.

Amusingly enough, the authentication failure *is* flagged as such within
the server, but on the way up it gets changed to a terminal value!!!

IIRC, we (I work for another computer company of three letters that
starts with 'S') have "fixed" this in our enhanced NFS server, but I
can't recall if we have pushed the changes up-stream yet, or if they
have been accepted... No, I take that back, I think we *hacked* in some
changes and sniffed the data stream to prove what was going on and then
put the bug on the "to be fixed" list, which is, ahem, rather long. IOW
- don't hold your breath waiting for it to be fixed!

Cheers,
Gary B-)

--
__________________________________________________ ____________________________
Armful of chairs: Something some people would not know
whether you were up them with or not
- Barry Humphries