Le 08-11-2007, à propos de
Re: RH linux NFS server; solaris 9 clients hang,
Jay G. Scott écrivait dans comp.unix.solaris :
> In article <slrnfj5r9f.ihi.knatschke@fermat.systella.fr>,
> JKB <wilhelm-siegfried.knatschke-koenigsberg@chezmoi.com> wrote:
>>Le 07-11-2007, à propos de
>>RH linux NFS server; solaris 9 clients hang,
>> Jay G. Scott écrivait dans comp.unix.solaris :
>>>
>>>
>>>
>>> they're making me move from a solaris file server to a red hat file
>>> server. red hat enterprise 4AS.U5. solaris 9 and solaris 10 clients.
>>> the solaris 10 clients have had the /etc/default/nfs
>>> NFS_CLIENT_VERSMAX=3
>>> set. that cleared up the initial connection issue.
>>> there is no such option in the solaris 9 file.
>>>
>>> what i see now is that, for long stretches, the RH server
>>> works fine. but then something happens and the solaris clients
>>> hang. i have notes about a 5 minute hang, but they seem to range
>>> from a few secs to, i dunno, 8 minutes.
>>> sorry, this is the solaris 9 clients hanging.
>>> the 10 clients aren't having problems that i notice.
>>> when the 10 clients failed they didn't work at all. now that they
>>> work a little, working only a little might be good enough that i
>>> never notice when they don't. but i bet they're okay. i've looked at
>>> the logs and i can't decide either way.
>>>
>>> i got a suggestion that it might be a buffer size issue.
>>> i tried wsize == rsize = 8K (which is supposed to be "low enough"
>>> ie, everybody can handle 8KB buffers) on the mount command.
>>> that __seems to have helped__ but it has __not cured__ the problem.
>>>
>>> so.... are there known problems between RH NFS servers and solaris 9
>>> clients? possibly solaris 10 clients?
>>
>> What is your kernel ? What is your nfs server ?
>
> the kernel is 2.6.9-55.0.9.ELsmp

Please change your kernel. Some bugs have been fixed since 2.6.20.
Some workarounds have been added too since your kernel.
2.6.23 works fine for me in your configuration (only one difference,
I only use sparc64 nfs server). Warning: if you has to use software
raid5, you have to wait 2.6.23.2 or 3. We have fixed a very
important bug in raid5 subsystem.

2.6.9 kernel is a very old and unmaintained kernel.

Regards,

JKB

--
Le cerveau, c'est un véritable scandale écologique. Il représente 2% de notre
masse corporelle, mais disperse à lui seul 25% de l'énergie que nous
consommons tous les jours.

JKB wrote:
> Le 08-11-2007, à propos de
> Re: RH linux NFS server; solaris 9 clients hang,
> Jay G. Scott écrivait dans comp.unix.solaris :
>> In article <slrnfj5r9f.ihi.knatschke@fermat.systella.fr>,
>> JKB <wilhelm-siegfried.knatschke-koenigsberg@chezmoi.com> wrote:
>>> Le 07-11-2007, à propos de
>>> RH linux NFS server; solaris 9 clients hang,
>>> Jay G. Scott écrivait dans comp.unix.solaris :
>>>>
>>>>
>>>> they're making me move from a solaris file server to a red hat file
>>>> server. red hat enterprise 4AS.U5. solaris 9 and solaris 10 clients.
>>>> the solaris 10 clients have had the /etc/default/nfs
>>>> NFS_CLIENT_VERSMAX=3
>>>> set. that cleared up the initial connection issue.
>>>> there is no such option in the solaris 9 file.
>>>>
>>>> what i see now is that, for long stretches, the RH server
>>>> works fine. but then something happens and the solaris clients
>>>> hang. i have notes about a 5 minute hang, but they seem to range
>>>> from a few secs to, i dunno, 8 minutes.
>>>> sorry, this is the solaris 9 clients hanging.
>>>> the 10 clients aren't having problems that i notice.
>>>> when the 10 clients failed they didn't work at all. now that they
>>>> work a little, working only a little might be good enough that i
>>>> never notice when they don't. but i bet they're okay. i've looked at
>>>> the logs and i can't decide either way.
>>>>
>>>> i got a suggestion that it might be a buffer size issue.
>>>> i tried wsize == rsize = 8K (which is supposed to be "low enough"
>>>> ie, everybody can handle 8KB buffers) on the mount command.
>>>> that __seems to have helped__ but it has __not cured__ the problem.
>>>>
>>>> so.... are there known problems between RH NFS servers and solaris 9
>>>> clients? possibly solaris 10 clients?
>>> What is your kernel ? What is your nfs server ?
>> the kernel is 2.6.9-55.0.9.ELsmp
>
> Please change your kernel.

It's RHES (or rather RHAS).
If it's "Update 5", it's already pretty recent.
Don't know what the latest minor-revision is that can be had for 4U5.
Is this system registered and up2date?

> Some bugs have been fixed since 2.6.20.
> Some workarounds have been added too since your kernel.
> 2.6.23 works fine for me in your configuration (only one difference,
> I only use sparc64 nfs server). Warning: if you has to use software
> raid5, you have to wait 2.6.23.2 or 3. We have fixed a very
> important bug in raid5 subsystem.
>
> 2.6.9 kernel is a very old and unmaintained kernel.

No it isn't.
OK, it is, but not in this specific case.
RHAT backports fixes from the current kernel to their enterprise-kernel(s).

You just have to figure out (by pestering RHAT-support) which ones
actually applied to the enterprise-kernel in the first place.

Open a ticket with RHAT, it's a RHAT-problem, not a Solaris-problem
actually (it worked before with Solaris as server, right <g>?).



cheers,
Rainer