Segmentation fault during peer startup in failover configuration

[Matt Kassawara]

Configuration:

I'm running redundant servers using failover peer mode in which both daemons actively handle DHCP requests and share lease information.

Expected operation:

Under normal operating conditions, both daemons should start in any order, synchronize lease information, and begin handling DHCP requests.

What actually happens:

Using the Jaunty DHCP server package (dhcp3-server 3.1.1-5ubuntu8), the first daemon consistently starts up properly, but the second generally throws a segmentation fault. After several attempts, the second daemon eventually starts up properly. However, upon restart will resume the pattern of throwing segmentation faults. Once both daemons successfully start up, they seem stable and perform as expected.

Release information:

$ lsb_release -rd
Description: Ubuntu jaunty (development branch)
Release: 9.04

$ apt-cache policy dhcp3-server-ldap
dhcp3-server-ldap:
  Installed: 3.1.1-5ubuntu8
  Candidate: 3.1.1-5ubuntu8
  Version table:
 *** 3.1.1-5ubuntu8 0
        500 http://us.archive.ubuntu.com jaunty/universe Packages
        100 /var/lib/dpkg/status

Crash output:

See attachment for gdb output from package including debugging symbols.

[H.-Dirk Schmitt (dirk-computer42)]

Here the dhcpd also segfaults after start.

Jun 16 09:18:37 garfield dhcpd: WARNING: Host declarations are global. They are not limited to the scope you declared them in.
Jun 16 09:18:37 garfield dhcpd: Wrote 0 deleted host decls to leases file.
Jun 16 09:18:37 garfield dhcpd: Wrote 0 new dynamic host decls to leases file.
Jun 16 09:18:37 garfield dhcpd: Wrote 314 leases to leases file.
Jun 16 09:18:37 garfield dhcpd: failover peer dhcp-failover: I move from normal to startup
Jun 16 09:18:37 garfield kernel: [75173.137232] dhcpd3[9766]: segfault at 20 ip 0000000000000020 sp 00007fff8d769b78 error 14

[Chuck Short (zulcss)]

Hi,

Can you please post your configuration file so I can try to recreate the bug?

Thanks
chuck

[Matt Kassawara (ionosphere80)]

See attached configuration files for primary and secondary servers.

[Rob Shinn (rob-shinn)]

I'm seeing the more or less the same thing: whichever DHCP server daemon was started second is the one that randomly segfaults during failover peer sync. My failover configuration is pretty much identical, but I'm not using LDAP to store leases.

[Istvan Botka (boti)]

Check time on both server. I had some kind of problem (segfault on master when peer communication starts). I've found theese messages in the syslog of slave.:

dhcpd: failover: disconnect: time offset too large

I`ve set up ntp on both server and dhcp cluster gets run fine.

[Chuck Short (zulcss)]

Thanks for the response.

Regards
chuck