Samba has stopped logging!

[Tony Walker]

Hi Guys,

I have a samba share on a 11.00 machine which is in regular use. On the 15th, a problem with one of the apps caused the filetable to fill and ultimately the machine to crash. The smb.conf is set so that a log is created per user (i.e /opt/samba/var/log.%U) but no user log has been updated since the crash. The last file to be updated is simply called "log." and the last line is:

[2002/10/15 20:38:24, 0] smbd/server.c:(220)
open_sockets: accept: File table overflow

Any help would be much appreciated.

Thanks,

Ton

[Ceesjan van Hattum]

users or no users, samba always want to log in case logging is turned on.
A file table overflow means that you have reached the maximum number of open files all over your system. This might have been cause by Samba, but also by other processes.
If you want to keep all processes (including their logging) just like it is, than you should reconsider the Configurable Kernel Paramters (change them by using 'sam').

Look for the variable 'nfile'.
This is not just a number, but should be a formula, containinig other variables.
It is to you to change or not to change certain other numbers and/or variables.

You can find a lot of information about in the internet using 'google.com'.

Remeber: if you change kernel variables, you will need to reboot the system to let the changes take effect.

Regards,
Ceesjan

[Tony Walker]

Ceesjan,

The filetable overflow was cause by another application an the problem has since been resolved. I have also increased 'maxusers' - thus increasing a variety of other formula based parameters including 'nfile'.

Thanks,

Tony

[T G Manikandan]

http://www2.itrc.hp.com/service/cki/docDisplay.do?docId=200000006417730

Increase MAXUSERS which in turn increases the nfile and ninode parameters.


Thanks

[T G Manikandan]

Did you check up the smbd.log and nmbd.log files for any hint.

You can the restart of the ambd daemon to force to re-read the smb.conf file.

Thanks

[T G Manikandan]

smbd not ambd .sorry

[John Warminski]

You didn't say, but my guess is that logging in general may have crashed .

Have tried a kill HUP to the logging daemon ... or stop and restart logging.

[Tony Walker]

Well, after speaking with one of my guys here - he made some changes to the configuration (local and domain master set to 'NO') and it suddenly started logging again! Will probably never know exactly what fixed it but there you go. Thanks for the replies.

Tony