dhcpd stops issuing leases with sporadic "no available billing: lease limit reached in all matching classes"

Richard Laager rlaager at wiktel.com
Fri May 7 06:07:40 UTC 2010


Every day or so, our dhcpd starts logging "no available billing: lease
limit reached in all matching classes". It'll do this for varying
lengths of time... I believe we've seen 30 minutes to 6 hours. Customers
on certain networks will effectively see DHCP as "down" (i.e. they don't
get leases). Other networks served by the same instance of dhcpd,
however, get leases fine during this time.

When this first started happening, we didn't catch it right away and it
would sometimes start working again on its own. When we did catch it,
we'd restart dhcpd and it'll immediately work again. Now, we have a
watchdog script running from cron every minute to tail the logs and
restart dhcpd (when this error message shows up) to workaround this
issue.

We were running fine for several years on a system with Ubuntu "Dapper"
6.06 and what we believe was dhcpd 3.0.6 rebuilt from the dhcpd package
from Ubuntu "Hardy" 8.04. Why were we using a backport instead of the
dhcpd 3.0.3 that shipped with Dapper? We were having some sort of issue.
We think it may be the same as the issue we were having now, but it's
been at least a couple of years and nobody remembers for sure.

The old DHCP server suffered the "loss" of both hard drives
simultaneously. (They started returning lots of read errors and/or bad
data so the RAID card dropped one and the filesystem went read-only.) We
replaced the drives and installed Ubuntu "Lucid" 10.04, which shipped
with dhcpd 3.1.3. We were able to grab the conf and leases files from
the old server before replacing the drives. The new install appeared to
work fine, until a day or so later when this error started showing up.

We tried rebuilding the 3.0.6 package from Hardy, but that did not
resolve the issue. Then, we rebuilt the 4.1.1 package from Debian
experimental. That has the same issue.

I believe that 3.0.6 was using a more generic error (perhaps just "no
available billing"), so our watchdog script wouldn't trigger properly.
And we can't change the script to check for "no available billing", as
that seems to occur normally (when someone tries to get a lease when
they already have as many as they're allowed?).

I believe this is a bug as the behavior varies across a restart of
dhcpd. However, I can't be sure what is going on.

What should I do next?

Thanks,
Richard
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <https://lists.isc.org/pipermail/dhcp-users/attachments/20100507/c06b4d44/attachment.bin>


More information about the dhcp-users mailing list