Problems with samba (2.2.3&2.2.5)

[Dr. Peer-Joachim Koch]

Hi,
we want to use the latest (HP) version of samba on our file server, to get ACL working for
2000/XP.
However the current release (A1.09.->Samba 2.2.5) won't work and with the previous version we also have some trouble ....

In principal we are using the following smb.conf:
[global]
workgroup = BGC
netbios name = ORCA
netbios aliases = ATLAS RHEA BONITA
server string = DVA.b024.HPUX_Samba
interfaces = lan1
log file = /var/adm/samba/%m.log
socket options = TCP_NODELAY
local master = no
read only = no
case sensitive = no
preserve case = yes
short preserve case = yes
dos filetime resolution = yes
log level = 1
syslog = 3
oplock break wait time = 100 # new I changed this today
read size = 65535
password server = manta, andorra
security = server
printing = cups

Using the 2.2.5 version we get the following error message:
smbd/files.c:(215)
file_init: Information only: requested 10000 open files, 2038 are available.

I have no idea where are the 10000 open files are comming from.

Any idea ?

Second using 2.2.3a. we have
two major problems:
1) We get a couple of oplock errors leading to a lost session. (not seen under 2.0.7)
2) People are often log out. Somehow the authentification seems not to work (the PDC and BDC are working well) and the user account gets log out. The admin has to unlock the account under windows.
3)The access to the server is slow. We have ~150 users (113 smbd in the moment), and mainly XP users are waiting long (5-30sec) if the try to connect to a share. The server is only working as file server (nfs+samba) and the file systems are on 2 EMC raids. The l class is connected to the raids using 2 FC adapter
and 1 gigabit -> users.
The machine has 2 CPU+1GB RAM.
Most people are just using word or netscape (load ~0.1),
so I think the machine is not
the problem (under nfs it's fast).

Any hint's ??

Thanks for your help, Peer

[Steve Steel]

Hi

This can be caused by samba initialises badly due to bad netwrok setup at startup.

Check netstat -rn and the netconf file for things like gateways.

If the samba is connecting wrongly it could be counting badly.

I have seem something similar with cifs once
caused by a bad gateway


Steve Steel

[Steven E. Protter]

On the HP box.

look at:
/etc/rc.config.d/netconf

Make sure al external IP addresses in the setup like dns servers and gateways and ping them.

Do the same thin on the XP box.

Once you have that done, you should make sure the domain in smb.conf on both machines is the same.

Any discrepencies, make changes, and restart server or samba depending on what kind of problem it is.

On UX, to stop and start samba

/sbin/init.d/samba stop
/sbin/init.d/samba start

P

[Dr. Peer-Joachim Koch]

Hi,

the network seems to work great.
The gigabit is fast and the nfs
server never had a problem.

bash-2.03# netstat -rn
Routing tables
Dest/Netmask Gateway Flags Refs Use Interface Pmtu
127.0.0.1 127.0.0.1 UH 0 3532653 lo0 4136
141.5.16.22 141.5.16.22 UH 0 1879196 lan1 4136
10.0.0.7 10.0.0.7 UH 0 56 lan0:1 4136
141.5.16.12 141.5.16.12 UH 0 0 lan0 4136
141.5.16.0 141.5.16.22 U 2 0 lan1 1500
141.5.16.0 141.5.16.12 U 3 0 lan0 1500
10.0.0.0 10.0.0.7 U 3 0 lan0:1 1500
127.0.0.0 127.0.0.1 U 0 0 lo0 4136
default 141.5.16.1 UG 0 0 lan1 1500

(lan0 is a second 100MBit interface).

Samba is only using lan1 (see smb.conf).
Yesterday I installed on a second l-class (new installed) samba 2.2.5 (A.1.0.9) and it runs perfectly (domain member,
no oplock errors ...).

Might the kernel be the problem ?
How big should maxuser be on
a NFS/samba file server ??
Is each smbd one user (as far as i know something like a "su $uid" is done from samba).

However the network is running well...

Bye, Peer

[U.SivaKumar_2]

Hi,

These kernel parameters should be modifed

nfiles = 10000

nflocks = 10 * Maximum number of Clients

Check the nfiles usage with
#sar -v 1 10

regards,

U.SivaKumar

[U.SivaKumar_2]

Hi,

Recommeded value for nfiles parameter for samba is nfiles = (23 + max_client_files) * max_connected_clients

regards,
U.SivaKumar

[Dr. Peer-Joachim Koch]

Hi,

our file server has the following values:
ash-2.03# sar -v 1 10

HP-UX orca B.11.00 U 9000/800 02/06/03

14:17:34 text-sz ov proc-sz ov inod-sz ov file-sz ov
14:17:35 N/A N/A 214/2048 0 1876/2728 0 3275/41618 0
14:17:36 N/A N/A 214/2048 0 1875/2728 0 3273/41618 0

What does inod-sz mean ??

It look's like there is enough room for samba 2.2.5 ... !?

bye, Peer

[Andi Rigauer]

one small thing maybe it helps you.
have u ever tried the webmin interface.
www.webmin.com?
its a nice tool to administrate your whole unix box also sambe and acl's
brgds
ANdi