[Lustre-discuss] serious samba+lustre problem

[Papp Tamás]

Dear All,

We have a really serious problem.

This weekend we upgraded our mixed 1.6.0.1 + 1.6.4.1 cluster based on 
CentOS 4.4 i386 to CentOS 5.1 x86_64.
We reformatted all drbd storages, so recreated the whole cluster and 
copy back from backup.

Windows clients reached the cluster through a samba server, but now it's 
just not working at all.


This is my smb.conf

# testparm
Load smb config files from /etc/samba/smb.conf
Processing section "[cluster]"
Loaded services file OK.
Server role: ROLE_STANDALONE
Press enter to see a dump of your service definitions

# Global parameters
[global]
        server string = Cluster gateway (SAMBA server)
        interfaces = 192.168.0.71/24
        log file = /var/log/samba/%m.log
        max log size = 50
        max xmit = 262144
        deadtime = 15
        kernel change notify = No
        socket options = TCP_NODELAY IPTOS_LOWDELAY SO_RCVBUF=8192 
SO_SNDBUF=8192 SO_KEEPALIVE
        load printers = No
        preferred master = No
        local master = No
        domain master = Yes
        dns proxy = No
        wins server = 192.168.0.100
        host msdfs = Yes
        hosts allow = 192.168.0.
        block size = 16384
        write cache size = 16777216

[cluster]
        path = /mnt/lustre
        valid users = user
        read only = No
        create mask = 0775
        guest ok = Yes
        map system = Yes
        oplocks = No
        level2 oplocks = No

This is on CentOS 4.4, but also I tried it on Debian 4.0 and default 
config with the same results.

Symptomps:

Sometimes I don't see anything in the share, sometimes, after samba 
restart, I see a directory, but only for the next refresh. If I change 
the mountpoint to user:user I see this in logs, otherway nothing 
usually, but not always and not always after that:

[2008/04/06 23:51:25, 0] smbd/service.c:make_connection_snum(620) 
'/mnt/lustre' does not exist or is not a directory, when connecting to 
[cluster]

Lustre is now 1.6.4.3 x86_64

I appreciate any help.

Thank you,

tamas