Subj : Testing Samba again
To   : All
From : Digital Man
Date : Wed Feb 16 2005 05:03 pm

  Re: Testing Samba again
  By: Digital Man to All on Wed Feb 16 2005 04:40 am

 > Due to the utter failure of smbfs or cifs to work correctly on any Linux or
 > FreeBSD system I can manage to put together, I'm trying once again to use my
 > Linux box as the central file server (using Samba v3.x this time).
 > 
 > I tested a couple years ago using Samba v2.2 and it was very noticeably slow
 > than Win2K on the exact same hardware.
 > 
 > So far so good with Samba v3.x. And this will once again give me some *nix B
 > nodes to play with (hopefully). :-)

smbd (3.0.8) went haywire and CPU bound after just a couple of hours.

There were a number of errors in the syslog, so I upgraded to 3.0.10 (the
latest RPM for FC3) and reduced the local load on the share directories (for
now).

I was still getting lots of "oplock" related errors in the syslog so I
investigated oplocks further and determined they are a very bad idea for a
Synchronet (or any active database) system. So I disabled them.

Things are running smoothely for now.

There are 2 instances of smbd:

 2765 root      15   0 12200 3788 9288 S 45.0  1.0  87:33.23 smbd
 2751 root      16   0 12256 3860 9288 S 36.7  1.0 193:45.19 smbd
                                         ^^^^ CPU utilization

at least one instance is at *least* 20% or more at any given time. Doesn't this
seem rather high?

The system is a PII-300, so it's no speed daemon, but the identically equipped
Win2K server never had such high CPU utilization while sitting idle (just
serving files).

                                                  digital man

Snapple "Real Fact" #108:
Licking a stamp burns 10 calories.

.