DHCP cluster crashes after a few hours
Describe the bug
When running a cluster using dhcpd 4.4.1 or 4.4.2, at least with Ubuntu patches, dhcpd crashes after a few hours.
Here are two instances of the crash with the packaged 4.4.1 from Ubuntu 20.04:
2020-07-31T06:28:28.138646-05:00 salmon sh[764]: ../../../../lib/isc/unix/socket.c:3361: INSIST(!sock->pending_send) failed, back trace
2020-07-31T06:28:28.138704-05:00 salmon sh[764]: #0 0x7fdd3f4b4a4a in ??
2020-07-31T06:28:28.138768-05:00 salmon sh[764]: #1 0x7fdd3f4b4980 in ??
2020-07-31T06:28:28.138809-05:00 salmon sh[764]: #2 0x7fdd3f4f07e1 in ??
2020-07-31T06:28:28.138849-05:00 salmon sh[764]: #3 0x7fdd3f297609 in ??
2020-07-31T06:28:28.138887-05:00 salmon sh[764]: #4 0x7fdd3f3d3103 in ??
2020-07-31T07:02:54.013649-05:00 salmon sh[32432]: ../../../../lib/isc/unix/socket.c:3361: INSIST(!sock->pending_send) failed, back trace
2020-07-31T07:02:54.013674-05:00 salmon sh[32432]: #0 0x7fb12a7e0a4a in ??
2020-07-31T07:02:54.013693-05:00 salmon sh[32432]: #1 0x7fb12a7e0980 in ??
2020-07-31T07:02:54.013711-05:00 salmon sh[32432]: #2 0x7fb12a81c7e1 in ??
2020-07-31T07:02:54.013728-05:00 salmon sh[32432]: #3 0x7fb12a5c3609 in ??
2020-07-31T07:02:54.013753-05:00 salmon sh[32432]: #4 0x7fb12a6ff103 in ??
To Reproduce
At this point, I'm not certain that it has to be a cluster configuration, but everyone reporting it (including me) seems to be running a cluster.
It's also not clear how much configuration is relevant either.
For me, it crashes within a couple of hours on the secondary system.
Expected behavior
dhcpd does not crash.
Environment:
- ISC DHCP version: 4.4.1 as packaged by Ubuntu (4.4.1-2.1ubuntu5) or 4.4.2 with the same Ubuntu patches as 4.4.1-2.1ubuntu5
- OS: Ubuntu 20.04 x64
The version from 18.04, 4.3.5-3ubuntu7.1, is fine.
Additional Information
https://bugs.launchpad.net/ubuntu/+source/isc-dhcp/+bug/1872118 and before that https://bugs.launchpad.net/ubuntu/+source/isc-dhcp/+bug/1870729
I am not the original reporter of either of those.
Some initial questions
Contacting you
Here is fine, or the Ubuntu bug, or rlaager@wiktel.com by email or XMPP. BTW, this item on the template references github, but you're now running your own GitLab instance, so that's probably old.