Andrew C. Dingman
adingman at cookgroup.com
Mon Sep 18 20:11:45 UTC 2006
I've got a what I think is a nice DHCP setup using a pair of RHEL 4 ES
64-bit servers and the Red Hat dhcp-3.0.1-54.EL4 package. They are a
failover pair serving several pools on different subnets through relay
agents. They are doing dynamic DHCP updates both to BIND using TSIG and
to an AD controller without authentication. They've been set up and
running like this for a couple months with no problems.
This Saturday, I added a subnet and it's attendant DNS zones, and we
fired up a new location. (Previously, this location had used their own
DHCP server on a class-c that happens to be in the middle of the home
office's class-b, and connectivity was maintained with proxy arp.)
Everything looked fine when I went home at 1:00 in the afternoon.
This Saturday at 20:33:34 (yes, they're using NTP) they both segfaulted.
Except for the memory address, the segfaults are identical.
Sep 16 20:33:34 keyx kernel: dhcpd: segfault at 000000562ade81a1
rip 000000552aad4787 rsp 0000007fbfffe860 error 4
Sep 16 20:33:34 alkyone kernel: dhcpd: segfault at
000000562ade8ee1 rip 000000552aad4787 rsp 0000007fbfffe860 error 4
Google turns up nothing more than a guy who had it happen on CentOS last
March every time he plugged in a certain D-Link device. Based on this
and the fact that my pair seem to have died at the same moment, I'm
guessing there was some sort of packet the daemon couldn't handle.
Is this a known issue? Is there a work-around, or some further data I
sould look for? Synchronized segfaults on the weekend don't make for a
More information about the dhcp-users