[21626] in Hotline Meeting
HP Printers Timing Out
daemon@ATHENA.MIT.EDU (Darrin E. Robinson)
Fri Mar 11 15:22:53 1994
To: op@MIT.EDU
Cc: probe@MIT.EDU, hotline@MIT.EDU
Date: Fri, 11 Mar 94 15:22:35 EST
From: Darrin E. Robinson <darrin@MIT.EDU>
I've looked into the 79 SERVICE problem and "hung queue" problems extensively
and may have found a fix.
TECHNICAL SUMMARY
The printers were closing the TCP port to the server causing the Transcript
filter (hpcomm) to die with an error of "Connection reset by peer." Talking
with Richard it was agreed that this was probably caused by some timeout
with the tcp socket connection. I first tried to change the timeouts on the
socket options on the server side to see if that would help. It did not. I
then found a parameter in the BOOTP Configuration file called idle-timeout.
# Idle timeout is the time (in seconds) after which an idle
# print data connection is closed. A value of zero disables
# the timeout mechanism. The default timeout is 90 seconds.
idle-timeout: 60
(I should shoot the collective HP Printer Division for not putting in
something like the TCP Data connection... this is NOT the same parameter
as the Front Panel's TIMEOUT option.)
We currently had this set to 60 seconds. So I turned up debugging in the
Transcript filter and timed the amount of time it took for the following
print command to finish and have the Transcript filter end;
athena% lpr2 -d -Phplj -z internetwork-mail-guide
It was taking longer than 60 seconds from the debugging information
"psif: done sending"
to the error
"/var/hp.test/pscommhp [main]: Error: Connection reset by peer"
So this shoued me that the idle-timeout was too low for printing this
particular file. Changing it to 0 (for indefinate) caused the job to
print sucessfully each time wihtout the "Fatal Error." I have timed this
job several times and it finished (after "done sending") in just over 60
seconds. I've changed the idle-timeout to 120 (2 minutes) on the printer
hero-p for testing. It works just fine.
END OF TECHNICAL SUMMARY
What this means is that I'd like to send a "reconfigure" to all the printers
sometime soon to change their TCP idle-timeout value to 120. This should
get rid of alot of "Fatal Error" and "daemon died" errors.
This in turn should clear up alot of the 79 SERVICE errors and mysterious
queue hanging problems.
Darrin Robinson
DCNS Distributed Systems Support
