[Samba] Fwd: samba & Oracle ACFS Issues

[Volker Lendecke]

On Tue, Dec 16, 2014 at 03:40:08PM +0100, Nacho del Rey wrote:
> Hi there
> 
> In an Oracle RAC cluster using ACFS (as file-system) where we have a samba
> server for sharing files to windows clients, we are suffering a strange
> issue, from time to time, which it causes the Windows clients lock for a
> while (10 min or a bit more) or even indefinitely when they are working
> with the share resource, and sometimes we have to re-start the samba
> service to come back to normality
> 
> We thought the problem was on the ACFS locks (incompatibility with samba
> locks), so we tried to avoid them using some parameters in smb.conf but
> unfortunately we are still having problems
> 
> Data:
> 
> OS: RHEL 6.4
> samba server: samba-3.6.9-151.el6_4.1.x86_64
> Oracle: 11.2
> Windows 2008 & Windows 7
> 
> smb.conf
> 
> [global]
>         log file = /var/log/samba/log.%m
>         socket options = TCP_NODELAY SO_RCVBUF=8192 SO_SNDBUF=8192

Please remove the SO_RCVBUF and SO_SNDBUF entries.

>         encrypt passwords = true
>         wins support = true
>         nt acl support = Yes
>         logon script = %U.bat
>         local master = yes
>         workgroup = SMBBD
>         name resolve order = bcast host
>         server string = Servidor Samba NOCON
>         security = share

Oh, that's ancient. Do you have any possibility to move to
"security=user"?

>         map to guest = Bad User
>         preferred master = yes
>         bind interfaces only = True
>         max log size = 50
>         domain logons = Yes
>         smb ports=139
>         log level = 2 locking:5
> 
> 
> [dossiers]
>         comment = xxxxxxxxxxxxxxxxxxxxxx
>         path = /opt/srv007/app
>         guest ok = yes
>         writable = yes
>         browseable = no
>         create mask = 0664
>         directory mask = 0775
>         force create mode = 0664
>         force directory mode = 0775
>         force user = expl
>         force group = expl
>         # The following to avoid file system lockings
>         locking = Yes
>         strictlocking = No
>         posixlocking = Yes

Did you try "posix locking = no"? That is mostly criticial
if you are exporting files from a file system with
unreliable locking like for example NFS.

>         oplocks = No
>         level2oplocks = No
>         fakeoplocks = No
>         blockinglocks = Yes
> 
> $  df /opt/srv007/app/
> Filesystem           1K-blocks      Used Available Use% Mounted on
> /dev/asm/vpdona07-390
>                       68157440  13112864  55044576  20% /opt/srv007/app
> 
> /dev/asm/vpdona07-390 /opt/srv007/app acfs rw,relatime,device,rootsuid 0 0
> 
> Has anybody got a clue?

Can you see what the smbd hosting such a blocked client does?
If it is in D state (according to ps u), it sits in the
kernel. If not, you could try stracing the process (strace
-ttT -p <pid>) and see what it does. gstack <pid> also helps
often.

With best regards,

Volker Lendecke

-- 
SerNet GmbH, Bahnhofsallee 1b, 37081 Göttingen
phone: +49-551-370000-0, fax: +49-551-370000-9
AG Göttingen, HRB 2816, GF: Dr. Johannes Loxen
http://www.sernet.de, mailto:kontakt at sernet.de