OK, I have been banging my head into the wall on this now for two or more
weeks
We have a samba server (2.2.8 then 2.2.8a) running on top of a 320 GB raid
ATA-raid array with a FastTrack controller. Both the server and workstations
are running through Gigabit NIC's. Every five or so minutes the
workstations
loose communications with the server, Then it takes 10-15 seconds for the
workstations to reconnect to the server. The times are not specific, for
example it can take anywhere from 1 minute to 3 hours between drop outs, but
when we put a workstation under load it is (approx) 5-7 minutes between drop
outs. Sometimes all stations loose the connection at once, sometimes it is
only one workstation. It does not occur in any other program except Samba.
If I don't get this fixed soon my boss is going to make me go back to
(shutters) windows
Globals section of the smb.conf file goes something like
---
#Server Identification
workgroup = GENESIS
netbios name = EXODUS
netbios aliases = INSTALLERS
server string = Fedler studio Samba Server
#Server Settings
log file = /data/system/samba/logs/%m.log
log level = 1
max log size = 5000
OS level = 99
#Security Settings
security = user
smb passwd file = /data/system/samba/smbpasswd
encrypt passwords = yes
pam password change = yes
#Networking settings
# interfaces = 10.0.0.16/24
# dns proxy = no
wins support = yes
preferred master = yes
#Domain Controller propertys
domain logons = yes
logon drive = y:
logon home = \\EXODUS\%U
logon script = logon.bat
domain admin group = @dAdmin
[netlogon]
path = /data/system/samba/netlogon
comment = network Logon Share
guest ok = yes
writable = no
share modes = no
---
Any help whatsoever would be appreciated
Andrew
On Thu, 29 May 2003, Andrew Holland wrote:> OK, I have been banging my head into the wall on this now for two or more > weeks > > We have a samba server (2.2.8 then 2.2.8a) running on top of a 320 GB raid > ATA-raid array with a FastTrack controller. Both the server and workstations > are running through Gigabit NIC's. Every five or so minutes the workstations > loose communications with the server, Then it takes 10-15 seconds for the > workstations to reconnect to the server. The times are not specific, for > example it can take anywhere from 1 minute to 3 hours between drop outs, but > when we put a workstation under load it is (approx) 5-7 minutes between drop > outs. Sometimes all stations loose the connection at once, sometimes it is > only one workstation. It does not occur in any other program except Samba. > If I don't get this fixed soon my boss is going to make me go back to > (shutters) windowsThis sounds VERY suspiciously like a network hardware problem. What NICs? What HUBs? Are you running in duplex mode? What error rates on the ethernet interfaces? What collision activity on the HUBs? This does not sound like a Samba problem. - John T. -- John H Terpstra Email: jht@samba.org
On Thu, May 29, 2003 at 10:47:16PM -0500, Andrew Holland wrote:> > We have a samba server (2.2.8 then 2.2.8a) running on top of a 320 GB raid > ATA-raid array with a FastTrack controller. Both the server and workstations > are running through Gigabit NIC's. Every five or so minutes the workstations > loose communications with the server, Then it takes 10-15 seconds for the > workstations to reconnect to the server. The times are not specific, for > example it can take anywhere from 1 minute to 3 hours between drop outs, but > when we put a workstation under load it is (approx) 5-7 minutes between drop > outs. Sometimes all stations loose the connection at once, sometimes it is > only one workstation. It does not occur in any other program except Samba. > If I don't get this fixed soon my boss is going to make me go back to > (shutters) windowsCan you get an ethereal trace between a client and server when this happens ? This will help to debug the problem. Thanks, Jeremy.
On Sunday 01 June 2003 12:51 am, you wrote:> On Sat, May 31, 2003 at 10:30:05PM -0500, theSyClops wrote: > > I admit that I am a linux-server newbie... How do I do the ethereal > > trace, I have not had to do it before and I wouldent know what to look > > for or how to do it > > The linux box should have a command called 'ethereal'. Run as root, > and get it to capture packets on the relevent ethernet interface > and filter the capture to only capture packets from and to the > client. > > If you've ever used the Microsoft Netmon application, ethereal > should be very easy. If you haven't used a network sniffer before, > I suggest you read up on the subject. > > Jeremy.Ok, after a quick crash course in ethereal including compiling and installing it, Here is the capture, This log has been filtered by time before and after where the drop out occured as well as source and dest IP's (frame.time_relative >= 343 && frame.time_relative <= 347.2) && (ip.src =10.0.0.224 || ip.dst == 10.0.0.224) Then saved (this info here for making sure I didnt screw up the logs) The error occures @ time 2 seconds after beginning of capture attachment available upon request, the list wont let me post it and I dont have a reliable space to throw it up on the net Andrew
ooops, its veto oplocks option sorry for the post Bri- __________________________________ Do you Yahoo!? Yahoo! Calendar - Free online calendar with sync to Outlook(TM). http://calendar.yahoo.com