[Samba] Samba stalls

J. Pilfold-Bagwell jools at oss4all.plus.com
Mon Jul 7 16:01:16 GMT 2008


Hi all,

I have an HP Proliant 115 (Dual Core Opteron) server with 4GB of RAM set  
up as a PDC with an identical machine acting as BDC. The PDC has an LDAP  
backend and I'm planning to add an LDAP slave on the BDC. Both are running  
Redhat EL5 with Samba 3.0.28.

The problem I have though is that after a period of time, samba seems to  
stall. Initially, I put it down to the Opteron dual core timing drift bug  
as I was getting negative ping times but after updating Grub with the  
relevant entry and rebooting, that was resolved. One of the member servers  
on the domain loses touch with the DC's when they stall and will only  
reconnect after Winbind is restarted.

I ran Gkrellm in daemon mode and found that the NIC has little if any load  
on it and the CPU cores never get above 50%. The total amount of RAM in  
use (including X11) is 191MB which seems remarkably good. When I  
partitioned the drives, I set the system disk up with /var/lib/ldap as a  
seperate partition to /var, and /var is where the action seems to be. Once  
/var kicks into life, samba seems to stop responding to clients and if you  
attempt to join a machine to the domain or carry out any other action that  
requires authentication from the Domain Controller, it will often time  
out. The data partition upon which everyone's home directories are stored  
is a Terabyte RAID 5 array which hardy ever seems to get above 1-2 MB/s.

My question therefore is what could the DC be up to?  Is it buffering up  
and dumping the data to disk very slowly? Anyone else been here?

Thanks in advance,

Jools




-- 
Using Opera's revolutionary e-mail client: http://www.opera.com/mail/


More information about the samba mailing list