Hi

We have some problems with the stability of cifs on a SLES10SP2 / OES2SP1 two node cluster.

Is the server fresh restarted and the cluster volumes mouted, all users are able to logon and working with all cifs shares. But, as soon as a user connected to this cifs share, in the cifs.log appears many messages like the following:

CIFS[8737]: WARNING: UTILS: In GetStatFromPath stat returned error : 2
CIFS[8737]: ERROR: UTILS: [CIFS_FillNTCreateAndXInfo]:Stat failed for file
CIFS[8737]: WARNING: UTILS: In GetStatFromPath stat returned error : 2
CIFS[8737]: ERROR: UTILS: [CIFS_FillNTCreateAndXInfo]:Stat failed for file
CIFS[8737]: WARNING: UTILS: In GetStatFromPath stat returned error : 2
CIFS[8737]: ERROR: UTILS: [CIFS_FillNTCreateAndXInfo]:Stat failed for file
CIFS[8737]: WARNING: SMB: IPC - TransactionNetShareGetInfo errorCode = 0
CIFS[8737]: WARNING: CLUSTER: TransactNetShareEnum: Filtering out share, nbtServer = 114d800, shareServer = 93d400
CIFS[8737]: WARNING: CLUSTER: TransactNetShareEnum: Filtering out share, nbtServer = 114d800, shareServer = 93d400
CIFS[8737]: WARNING: CLUSTER: TransactNetShareEnum: Filtering out share, nbtServer = 114d800, shareServer = 93d400
CIFS[8737]: WARNING: SMB: IPC - TransactionNetShareGetInfo errorCode = 0

These messages are reporting for every user logon and sometimes if a user is browsing within the cifs share with the windows explorer.

Next, after cifs is running fine for hours or may some days, the users cannot logon anymore. Cifs.log ist reporting following:

CIFS[8759]: CRITICAL: AUTH: Failed to fetch credentials of User : cn=User1,ou=UsersAliases,ou=OU1,o=Company Nmasldap-error : 81 , ProxyUser- cn=cifsProxyUser-Server1,ou=Services,ou=OU1,o=Company . Check Configuration
Mar 4 22:23:54 Server1 CIFS[8759]: CRITICAL: AUTH: GetMD4Hash Failed.
Mar 4 22:23:54 Server1 CIFS[8759]: WARNING: CODIR: Failed to authenticate User : User1
Mar 4 22:23:54 Server1 CIFS[8759]: CRITICAL: AUTH: tempSES->nwLoginError is 81

Mostly only restarting the Server will solve the problem. Only restart cifs does not. Going offline with the cluster resource often ends with segmentation fault in cifsd.

Does anyone have an idea?

Regards
Michael