Samba 2.2.5 problem

[Sanjiv Sharma_1]

Hi,

I have one N4000/HP-UX 11.00/Samba 2.2.5.

I have some Win'2000 prof. PC which maps some samba share to it.

Till yesterday everything was fine but when I am trying to map the samba drive it popup an error:

"The mapped network drive could not be creaed because the following error has occured.
No nore connection can be made to this remote computer at this time because there are already as many connections as the computer can accept".

To some other PCs the error message are:

" The mapped netowrs drive couldnot be created because the following error has occured:
The remote computer is not available".

I can see about 18 nos smbd daemons are running.

What has gone wrong and what needs to be done?

Thanks,

[Steven E. Protter]

1) Is Samba actually working on the server.

ps -ef | grep smb

If you don't get any hits do this as root:

/sbin/init.d/samba start

If this does not fix it, you need to check networking.

Run ping tests from client to server and server to client, then traceroute/tracert tests.

If any of these tests fail you have a network connectivity issue.

It could be the lan card on the server. It could be a lan communication issue, if Cisco built your switch.

Change the port settins on the Cisco switch to the top rated speed of the NIC on the HP box. Don't use auto negotiate, it confuses many HP NICS.

Then boot the hp box

then as root

lanadmin -x lan# # being the number of the lan card.

Don't be surprised if the speed setting doesn't match what you have hardcoded on the Cisco switch. If that happens, hard code the settings in this file.

/etc/rc.config.d/hpbtlanconf

Here is an example.



HP_BTLAN_INTERFACE_NAME[0]=lan0
HP_BTLAN_STATION_ADDRESS[0]=
HP_BTLAN_SPEED[0]=10HD
HP_BTLAN_INTERFACE_NAME[1]=lan1
HP_BTLAN_STATION_ADDRESS[1]=
HP_BTLAN_SPEED[1]=100FD

Good luck and boot after changing these settings.

SEP

[Balaji N]

hi
as SEP pointed, is ur smbd running. do a kill -HUP and see if it works.

also can u please post any relevant error messages from the log files.

-balaji

[Sanjiv Sharma_1]

Hi,

Samba is running on the servers 15 clients are able to map the drive.
The problem is the 16th and above users.

Is there any parameters like the number of Samba connection?

Where can I look for the error?

Thanks,

[Sanjiv Sharma_1]

Hi,

samba is running. 15 connections are working fine. Problem starts when we try to map the samba drive through 16th or more connections.

Is there any parameter which takes care of the max. no. of connections?

Thanks,

[Darren Prior · Accepted solution]

Hi Raje,

Have a look through the Samba logs for any mention of failing to access locks. It's possible that you've run out of locks if you haven't tuned the kernel to run Samba.

You haven't mentioned if you've only recently seen this problem. Recent versions of Samba are more resource intensive.

How many users are you expecting to connect to this Samba server?

regards,

Darren.

[Paul Senior]

Hi,

You may find it useful to increase the level of debugging on smbd - especially if your samba smbd log (located wherever smb.conf specifies) currently doesn't acknowledge the problem you're experiencing.
This can be done by stopping smbd and restarting it, specifying debug level, e.g.

./smbd -d3

We setup samba recently and found level 3 debugging very useful.

regards,


Paul S.

[Geoff Wild]

You need to icrease some Kernel parameters:



Updated HP-UX System Configuration Requirements for Version A.01.08
-------------------------------------------------------------------
CIFS/9000 Server, version A.01.08, utilizes system resources differently
from previous versions. The following HP-UX kernel parameters must be
adjusted accordingly:

o NFILES - total number of open files per system

Each smbd process will initially open 23 files for internal use.
However, clients will open more files during a session. Therefore, as a
minimum, the number of nfiles should be:

nfiles = (23 + max_client_files) * max_connected_clients

NOTE: Previous CIFS Server versions used 8 files for internal use.

o NFLOCKS - total number of file locks per system

At least ten (10) file locks are allocated per smbd process for internal
use. Clients may need additional locks depending on the applications
used. Therefore, as a minimum, the number of nflocks should be:

nflocks = 10 * max_connected_clients

NOTE: Previous CIFS Server versions did not require explicit
configuration of this parameter.


These guidelines for minimum parameter values account for CIFS/9000
Server usage of system resources only. Other applications and system
processes may require further increases in the value of these
parameters.


This means that the default configuration of 200 nflocks on HP-UX will only
allow less than 20 PC clients (dependent on other application demands for
file locks) to connect to the CIFS server before no further locks will be
available.

This can be addressed by reconfiguring the kernel with an appropriate number
for the nflocks value and rebuilding a new kernel with that parameter.

Note that when using Glance to monitor the nflocks value, it has been seen
that when CIFS Server has generated this error, Glance will report a maximum
usage much lower than the system configured maximum. However all evidence
indicates that the CIFS server is reacting to an ENOLCK error received from
the operating system when making the fcntl() lock request.

Also check out:

http://www.docs.hp.com/hpux/pdf/B8725-90020.pdf


Rgds...Geoff

[Geoff Wild]

Did you try the kernel parameter changes? Just want ot know if that helped you or not.

Rgds...Geoff

[Gary Yu]

Hi,

You can user samba's admin interface SWAT, it's defaultly running on port 901, there're lots of Samba parameters you can tunning, include max-connections.

thanks,
Gary