SCO Openserver 5 disconect all the workstations

[segbanegas]

Hello I'm using SCO openserver release 5. I have a problem every 3 or 4 days the unix server disconect the windows terminals and the serial terminal, all looks normal in the unix console but I need to reebot the unix server to be able to work again. I check some messages in the log file like
telnetd [13258] Can't find user in protected password database.
nmbd[514] error could not determine network interfaces, you must use a interfaces config line.
lockd[419] out of stream resources.

 

[Wonboodoo]

That last "lockd" error looks like you might be running out of file handles. Something may be opening a lot of files and not releasing them. You can use "sar -v" (I think) to see the number of open files. There are probably other methods as well. I think you can increase the maximum number of open files with a kernel parameter change. That said, I don't think this would kick the users out. Could be a red herring.

When you say "all looks well on the serial console" are you able to login there?

Where are the serial terminal(s) connecting too? Serial ports on the server or are they coming in via a terminal server? If coming in via a terminal server and your serial console is able to login you may be running out of network resources. There is a command to see network resource utilization. Forget what it is, someone else will know. You can increase kernel parameters NSTREAM, NSTRPAGES, & NUMTIM to bump up the maximums.

 

[segbanegas]

The serial stations are connected to a digiportserver II 16. Those are like 12 stations. The pc's are using procomm plus, other are using multiview all those ove telnet. The stations and the pc's loose the connection, if I try to ping the unix server I don't have response. If I go to the unix server and try to ping a station or a pc nothing.
That's happening every 3 or 4 days.

Help please. and Thann you.

Pos. I'm don't have experience in Unix I'm giving my first steps and having this problems.

Thank you.

 

[aarne]

Hey

Try netstat -m to see if there are any fail's. I've notised that fail's in class6 meens that you have same problems in lan. Maybe somebody can tell yuo more about that.

t.aarne

 

[Wonboodoo]

One of my colleagues had a similar problem.

Run "netstat -m" look at the data at the end:
total configured streams memory:
streams memory in use:
maximum streams memory used:
Especially look at a time either close to failure or after the failure occurs (I assume you can still log into the console). If "in use" and/or "maximum" are close to the "total" then you are probably running out of streams.

If you are running out of streams you may need to increase some kernel parameters:
1) run "scoadmin"
2) choose "Hardware/Kernel Manager"
3) choose "Tune Parameters"
4) choose "Streams"
5) increase NSTREAMS, NSTRPAGES, & NUMTIM. Not sure if you're required to increase them all but it's what we did. On our system (currently) we have these values: NSTREAM=9000, NSTRPAGES=3500, & NUMTIM=1500.
6) you should be asked to rebuild the kernel. Answer "y" to everything
7) reboot

Good luck