User sessions time-out and no connection.

[Toby Ragsdale]

As of this AM, user sessions are timing out, and in some cases users have to open a telnet session several times to get a connection. Nothing seems apparently wrong. The network is behaving well, when the users are connected the Unix box is responding well. If I connect to the console, it will stop responding in the same fashion.

I don't know where to start, but looking for anything relating to the box itself. I just don't see any other network problems. Nothing has changed since Friday.

Thanks
TBR

[Steven E. Protter]

Check the parameter TMOUT

It may be timing out users.

You may not have enough pty devices defined for all your users.

You can spot the problem as follows:

tail -f /var/adm/syslog/syslog.log

You can fix it by increasing the number of devices as follows:
kernel parameters
npty
nstrpty

Quick way is to use Sam
Kernel
Configurable Parameters

bump these up, actions process new kernel, a boot will be required.

SEP

[Toby Ragsdale]

Getting this error in syslog

Mar 8 10:50:10 phoenix telnetd[2518]: getpid: peer died: Error 0

We have not added users, more connections etc.

Thanks.

[Toby Ragsdale]

npty=60
nstrpty=60
maxusers=32

We are only at ~20 users.

[Helen French]

Check your network parameters, especially 'tcp_keepalive_interval' in /etc/rc.config.d/nddconf file. This parameter controls the telnet time out values (60000=1 mnt.)

Also make sure that you have applied all recommended patches.

[Helen French]

This document might be helpful (TKB #KBRC00012873):

http://www2.itrc.hp.com/service/cki/docDisplay.do?docLocale=en_US&docId=200000069216208

[Toby Ragsdale]

Looked at the nddconf. All lines were rem out. I took the # off three lines;...

Well I cant even connect to tell you. The lines set the timout to 60000.

Again. Nothing was changed from last week to today. All was well, this morning people started to report that it took several trys to connect. They would open a telnet session, and it would NOT connect. On try three or so, it would. Once connected, they may stay conneccted for 1/2 hr, and the session time-out.

Keep it comming. This problem is getting worse.

Thanks
TBR

[Toby Ragsdale]

Just to add a note;

When I run top it starts with a normal 1/7 pages. After a min or so, it shows 60-120 pages! Whats up with that?

[Robert Salter]

The sudden changes in top could be the user sessions dying and attempts to re-connect. Or there's something running that's using your resources. Does the telnetd PID change in syslog.log?

[Toby Ragsdale]

Yes different PID.

[Toby Ragsdale]

All,

Thank-you ALL soo much. As usual I have received excellent help, while shooting myself in the foot.

Seems while working on a call recorder over the weekend, on reboot it reverted to a old address conflicting with my Unix server. Well... I guess that all things being equal, I would rather have it be a simple error on my part than anything else.

Thanks again.
TBR

P.S. On another Q I had trouble with the 'Points' system. I will check back to see that they get assigned.