Found in logs at the reported time ...
netrekd: hit maximum processes, connection closed
It seems we have a reaper leak in newstartd.
I was able to reproduce the symptom by connecting multiple observer
clients until the error appeared.
% ps ax|grep netrekd
591 ? S 11:06 ./netrekd
588 ? S 178:52 ./netrekd
% ps axl|grep 588|wc --lines
22
% gdb newstartd 588
(gdb) p active
$1 = 88
(gdb) set active=22
(gdb) p active
$1 = 22
(gdb) quit
Symptom has gone away.
The code was designed to prevent a denial of service attack, but one
could almost argue that it is providing a means to deny service.
Continuum's newstartd.c has "#define REAPER_HANDLER". I've restarted
netrekd with debug logging enabled. This will help me find the
circumstances where the active count becomes corrupted.
--
James Cameron mailto:quozl at us.netrek.org http://quozl.netrek.org/