CIFS Server problems

[James Millsap_2]

I currently run CIFS 1.09.01. I have 150+ X-4000 machines mounting shares. I have now had a problem where these X-class workstations cannot connect to the samba shares. I can connect with my toshiba laptop. To resolve this issue I restarted samba and suddenly everyone was able to get in. This is a bit confusing, any ideas.

HP RP-7400
HPUX 11.00
Jim

[Brian Bergstrand]

Did you check the smb/nmb logs in /var/opt/samba for any errors? What logging level are you using? Have you looked at any of the individual machine logs in the same dir to see if they contain information on the disconnect problem? Finally, what is the specific error on the client?

[Darren Prior]

Hi Jim,

The reply above has some good troubleshooting ideas. Here's some other points to consider:

If you've recently upgraded from a pre-A.01.08 version of CIFS or this is the first time you've run CIFS, have you increased the kernel parameters listed in the documentation? With the number of clients you have, it's likely that the demand on system resources will require kernel tuning. The fact that restarting CIFS resolved the problem does suggest that it is resource related.

I know very little about the X-class, except that they can run Linux or Windows - which specific OS and version are you running?

regards,

Darren

[James Millsap_2]

We have had 1.09 in production for about four months. Unfortunately all the logs get flushed when samba is restarted. The X-class machines are running W2K.

the error msg in syslog was this.

Aug 13 05:32:37 lgprime smbd[23518]: [2003/08/13 05:32:37, 0] lib/util_sock.c:(
436)
Aug 13 05:32:37 lgprime smbd[23518]: read_data: read failure for 4. Error = C
onnection reset by peer

sorry for the word wrap but you get the idea.

[Darren Prior]

Hi Jim,

I'd suggest that you increase the log level, if it's only at the default of 0 you get virtually no information. A log level of 2 should be sufficient to get further information if it happens again. It would also be a good idea to stop it from clearing the logs when it restarts, which is definitely not the default behaviour.

regards,

Darren.

[James Millsap_2]

hmm, ok

As far as samba goes this is out of the box. I installed it, modified smb.conf for the shares and domain info. thats about it. also as resources go we had a file locking problem which was fixed by uping the nflocks parm to 2000. this happenend after our oracle 9i implementation. Thats pretty much all the info I have. we were not having this problem until we went to 1.09. It seems to manefest itself every month or so.

Thanks in advance