[Samba] hanging smbd(s) revisited

[Matt Johnson]

On Tue, 28 Feb 2006, Fermin Molina wrote:

>>>> Just to add -- our fcntl locking issue is on Linux, we've
>>>> seen it on 2.6.9, 2.6.13.1 and 2.6.15.3, running Mandrake 10.2.
>>>> locking.tdb is on a local disk. All smbd child processes are
>>>> blocked on apparently the same fcntl when it happens.
>>>
>>> Hmmm...ok.  That ruins my theory.  I thought you were on AIX
>>> as well.  And just to make sure, you are running Samba 3.0.21b
>>> as well?
>>
>> Correct -- 3.0.21b on Linux 2.6. We had the same problem with 3.0.20 but
>> it was MUCH more frequent... 3.0.21b seems to have reduced the frequency
>> of the problem occurring but it does still seem to be there.
>
> I have had exactly the same problem. I reported that in thread "Samba
> daemons hang trying to lock locking.tdb", about Jan 25.
>
> I had a NFS mount from another server and then shared with samba. I
> moved the information on that server localy to the Samba server
> (avoiding NFS). All goes ok from that change.
>
> I don't understand why it works, because all smbd daemons were hanging
> in the previously commented fcntl call, that locks "locking.tdb" (that
> was located in local filesystem, not in any NFS mounted shares).
>
> I'm using FC4 with last updates (kernel included) and samba 3.0.21b.

That's the one. We are indeed sharing volumes from the Samba server 
which have been mounted via NFS, perhaps this is indeed an issue? 
Something subtle with regard to tdb locking deadlocking in very specific 
cases for samba-reshared NFS filesystems? (And yes, our locking.tdb is 
on a local /var.)

To reiterate... the fcntl64 spins in a blocking wait, so we never see an 
error unfortunately. It is always one of the 1-byte locks in locking.tdb 
which it jams on. Also, no change to this behaviour from 3.0.20 through 
3.0.21b -- all have been broken, although 3.0.20 seemed to break more 
frequently than 3.0.21b.

cheers

Matt

-- 
======================================================================
Matt Johnson <mwj at doc.ic.ac.uk>               (020) 7594 8440 / x48440
Systems Programmer, Computing Support Group         Office: Huxley 225
Department of Computing, Imperial College London
======================================================================