relayd dies lost child: socket relay engine terminated

[Marian Hettwer]

Hi Folks,

I'm having some trouble at work with a pair of OpenBSD 4.8's doing 
load balancing with relayd.
Pretty often, relayd dies. /var/log/syslog shows:
Feb 17 17:13:14 openlb38-2 relayd[24485]: check_child: lost child: 
socket relay engine terminated; signal 11

I'm helping myself right now with this beauty of a one-liner:

root@openlb38-2:~ # while sleep 1; do pgrep -u _relayd > /dev/null && 
continue; date ; echo restarted ; relayd; done
Thu Feb 17 17:13:15 CET 2011
restarted
Thu Feb 17 17:20:47 CET 2011
restarted
Thu Feb 17 17:23:43 CET 2011
restarted

As one can see, this counts as pretty often.

this is: OpenBSD 4.8 i386

relayd started acting like that one me after I added 3 groups for ssl 
offloading.
Like that (from relayd.conf)

relay cimobilelbssl {
       listen on $cimobilelb_addr port 443 ssl
       protocol "http_ssl"
       forward to <front38> port 80 check tcp mode loadbalance
}

relay ciebayklbssl {
       listen on $ciebayklb_addr port 443 ssl
       protocol "http_ssl"
       forward to <front38> port 80 check tcp mode loadbalance
}

relay wikilbssl {
       listen on $wikilb_addr port 443 ssl
       protocol "http_ssl"
       forward to <front38> port 80 check tcp mode loadbalance
}

I really love OpenBSD + carp + relayd, ... so I really like to help 
debugging this.
What datapoints do you need? Any hints?
Since relayd is failing pretty often, should I start it with ktrace 
for a closer look?
I couldn't find a core dump or anything like that after relayd fails.

Any help and pointers are highly appreciated :)

best regards,
Marian

PS.: keep me CC'd, as I'm not subscribed