telnetd[2068]: getpid: peer died: Error 0

[William Martinez Yanez]

The system kill some users, without reasons, the message in /var/adm/syslog/syslog.log is
telnetd[2068]: getpid: peer died: Error 0

This is new to me,

[Steven E. Protter · Accepted solution]

Shalom,

Are you sure the system killed the user?

The message tells me that the client disconnected. That could be a software problem with the telnet client, it could be a network problem, but it does not necessarily mean the server has a problem.

To be safe,

vi /etc/inetd.conf

Take a look at the parameters for telnetd

inetd -k
inetd

That will restart the Internet services daemon.

SEP

[A. Clay Stephenson]

Did you lose a network connection to a subnet (e.g loss of a router, WAN connection)? You might see this if the system is out of pty's so that npty and nstrpty might need to be inceased. (These tunables are among the few that SAM really handles well because not only are the kernel tunables increased but also the additional device nodes are automatically created.)

[Bill Hassell]

Actually, the HP-UX system did not kill these sessions -- the telnet protocol has been terminated by an interruption in service. An unreliable connection (firewall, ISP, etc) will cause these disconnects. When telnet loses the connection, it signals all processes with a SIGHUP (hangup signal) and the processes rightfully terminate since there is no more connection. This is a fairly common error message in 10.20 or earlier and unless you are missing patches, there will be no fix. For supported versions of HP-UX, you'll likely need to monitor the connections with nettl (or Wireshark) to see how the protocol is being trashed.