samba - Dec 2018 - Obtaining lock loop

Hello everyone,

I am working with a database shared with samba and sometimes the client and
server enter in a loop where the client try to obtain a lock to a file and
the server respond with "STATUS_LOCK_NOT_GRANTED". This situation can
last
until one of them crash.

I think in the past lock spin count prevent this from happening. In
nowadays samba version is there an option that replaces old lock spin count?

Thanks

On Wed, Dec 12, 2018 at 08:13:18PM -0300, Jose Concha via samba
wrote:
> Hello everyone,
> 
> I am working with a database shared with samba and sometimes the client and
> server enter in a loop where the client try to obtain a lock to a file and
> the server respond with "STATUS_LOCK_NOT_GRANTED". This situation
can last
> until one of them crash.
This can happen for several different reasons.
> I think in the past lock spin count prevent this from happening. In
> nowadays samba version is there an option that replaces old lock spin
count?
Not for SMB2+, it's not needed. We need to see a wireshark
trace and debug level 10 from smbd showing what is going on in order to debug.

I have a wireshark trace, which I attach. I will try to get a level 10 log.

Thanks

El mié., 12 dic. 2018 a las 21:44, Jeremy Allison (<jra at samba.org>)
escribió:
> On Wed, Dec 12, 2018 at 08:13:18PM -0300, Jose Concha via samba wrote:
> > Hello everyone,
> >
> > I am working with a database shared with samba and sometimes the
client
> and
> > server enter in a loop where the client try to obtain a lock to a file
> and
> > the server respond with "STATUS_LOCK_NOT_GRANTED". This
situation can
> last
> > until one of them crash.
>
> This can happen for several different reasons.
>
> > I think in the past lock spin count prevent this from happening. In
> > nowadays samba version is there an option that replaces old lock spin
> count?
>
> Not for SMB2+, it's not needed. We need to see a wireshark
> trace and debug level 10 from smbd showing what is going on in order to
> debug.
>