[syslog-ng] mysterious kill -1 death

[Joakim Rastberg]

Hi list!

I have a strange fenomen to report:

Sometimes when doing "kill -1" on a client that is sending logs with tcp
to a server, the _server_ dies silently. Truss gives a *big* burst of
output on the server (about 300 lines of time() calls), it looks like all
the hardware devices (keyboard, mouse, stdin, disks, nw-cards) outputs a
line into kern at the same time.

I cannot reproduce the death on low-volume machines, only on our
prime syslog server (100 hosts, about 5 messages/second) dies sometimes
when a client with syslog-ng (with tcp enabled) does a "kill -1".

Have anyone experienced anything like this?

syslog-ng 1.4

CLIENT:
SunOS unna4 5.6 Generic_105181-19 sun4u sparc SUNW,Ultra-2

options { long_hostnames(off); sync(0); use_fqdn(yes);};
destination loghost { tcp("hostname.domain.tld" port(514)); };
log {.... destination(loghost); };

SERVER:
SunOS syslog 5.7 Generic_106541-09 sun4m sparc SUNW,SPARCstation-20

options { long_hostnames(off); sync(0); use_fqdn(yes); use_dns(yes);};
source tcplog { tcp(ip(0.0.0.0) port(514)); };
destination tcpauto { file("/var/log/tcp/$HOST.$FACILITY"); };
log { source(tcplog); destination(tcpauto); };

/Joakim