Hi Ulrich,

I can try to reproduce this on my systems but I have a couple of questions
for you.

This is obviously an 11i v3 system since you're using NFS v4. Does this
problem only happen with NFS v4? If you do a loopback NFS mount using NFS
v3 does it work as expected? What version of the ONCplus ISU is installed
on the system? If you're not already running the B.11.31.04 ISU release you
might want to install that just to make sure you're running the latest
version of ONCplus.

Also, are you using some kind of security flavor other than sys in your
environment such as KRB5? It would help me if you could provide the exact
syntax of the "share" command so I can see exactly how the local filesystem
is being shared. I'd also like to see the "nfsstat -m" output so I can see
exactly how the loopback NFS filesystem is mounted. If you could also
provide the output of the "quota -v -u user" syntax so I can see how the
quotas are configured for the user in question.

Thanks,

Dave



"Ulrich Windl" wrote in message
news:87ej4dfny6.fsf@pc9454.klinik.uni-regensburg.de...
> Hi!
>
> when loop-mounting (i.e. via localhost) a VxFS 5 filesystem with quotas
> vie
> NFSv4, syslog reports "vxfs: WARNING: msgcnt 377532 mesg 089: V-2-89:
> quotas
> on /export/filesys invalid; disk usage for user id 123 exceeds . sectors"
>
> Now the thing is that the user reported has no quota limits set (has all
> "0"
> values), and "quota -v" reports no quota for the user, and quotacheck
> finds no
> problem.
>
> Also when trying to copy a large file to the filesystem, cp first reports
> no
> error, but the file has zero length then. On a second attempt for the same
> file, copy reports an I/O error with a quota problem.
>
> Another, probably unrelated problem are messages in syslog saying "NOTICE:
> sec_clnt_geth: unknown authflavor ###, trying auth_NONE" where "###" is a
> varying number.
>
> Also we see an NFSv4 mountpoint with a mounted filesystem changes to a
> file
> with odd permissions and date when the NFS server is restarted. On the
> client
> usually an umount/mount helps to let the directory re-appear.
>
> Can anybody explain what's going on?
>
> Regards,
> Ulrich

"Dave Olker" writes:

> Hi Ulrich,
>
> I can try to reproduce this on my systems but I have a couple of questions for
> you.
>
> This is obviously an 11i v3 system since you're using NFS v4. Does this
> problem only happen with NFS v4? If you do a loopback NFS mount using NFS
> v3

Hi!

About three weeks before we had moved to HP-UX 11.31 with the latest
ONCplus. The filesystem had been exported then over ethernet to PA-RISC HP-UX
11.11 (effectively using NVFv3). When moving the "client" to the same host as
the server (a SG packet), NFSv4 was used. A few hours after that those stream
of error messages started.

> does it work as expected? What version of the ONCplus ISU is installed on the
> system? If you're not already running the B.11.31.04 ISU release you might

Loopback mounts usually work as expected, but I have an odd feeling if on the
machine a user accesses the same file system via the locally-mounted &
exported filesystem as well as via the NFS-loopback-mounted directory (typical
for HA-NFS).

B.11.31.04 (latest, two versions newer than the last official OE)

> want to install that just to make sure you're running the latest version of
> ONCplus.
>
> Also, are you using some kind of security flavor other than sys in your
> environment such as KRB5? It would help me if you could provide the exact
> syntax of the "share" command so I can see exactly how the local filesystem is
> being shared. I'd also like to see the "nfsstat -m" output so I can see

Sharing is doene via the ServiceGuard NFS package. /etc/dfs/sharetab uses this
syntax:
/filesys - nfs root=host1,rw=host1:host2:host3:...

The mount options are quite standard (nfs4,tcp,hard,sec=sys,intr,32kB read
write size)

> exactly how the loopback NFS filesystem is mounted. If you could also provide
> the output of the "quota -v -u user" syntax so I can see how the quotas are
> configured for the user in question.

"quota -v -u user" produces just a heading line

I hope this is hwat you need. If you want to have more details, please contact
me per E-Mail.

Regards,
Ulrich


>
> Thanks,
>
> Dave
>
>
>
> "Ulrich Windl" wrote in message
> news:87ej4dfny6.fsf@pc9454.klinik.uni-regensburg.de...
>> Hi!
>>
>> when loop-mounting (i.e. via localhost) a VxFS 5 filesystem with quotas vie
>> NFSv4, syslog reports "vxfs: WARNING: msgcnt 377532 mesg 089: V-2-89: quotas
>> on /export/filesys invalid; disk usage for user id 123 exceeds . sectors"
>>
>> Now the thing is that the user reported has no quota limits set (has all "0"
>> values), and "quota -v" reports no quota for the user, and quotacheck finds
>> no
>> problem.
>>
>> Also when trying to copy a large file to the filesystem, cp first reports no
>> error, but the file has zero length then. On a second attempt for the same
>> file, copy reports an I/O error with a quota problem.
>>
>> Another, probably unrelated problem are messages in syslog saying "NOTICE:
>> sec_clnt_geth: unknown authflavor ###, trying auth_NONE" where "###" is a
>> varying number.
>>
>> Also we see an NFSv4 mountpoint with a mounted filesystem changes to a file
>> with odd permissions and date when the NFS server is restarted. On the
>> client
>> usually an umount/mount helps to let the directory re-appear.
>>
>> Can anybody explain what's going on?
>>
>> Regards,
>> Ulrich

Ulrich Windl writes:

> Hi!
>
> when loop-mounting (i.e. via localhost) a VxFS 5 filesystem with quotas vie
> NFSv4, syslog reports "vxfs: WARNING: msgcnt 377532 mesg 089: V-2-89: quotas
> on /export/filesys invalid; disk usage for user id 123 exceeds . sectors"
[...]

Just to keep you updated: A reboot of the machine does not fix the problem.
As far as HP support is concerned: Level-1 support tried to tell me that it's
expected bahavior. When I objected, the call went to Level-2, which resigned
very quickly, so that now Level-3 is "working" on it. In practice that means
that there was no reasonable suggestion within 5 days...

Regards,
Ulrich