[Pdns-users] Recursor backend timeout

[Olly Butterfield]

Hi there

I'm having some problems on a nameserver machine which uses pdns as a
nameserver on one IP and bind running on another IP to act as a recursive
backend for pdns. Bind has been set up to provide recursion only, and all
queries come into pdns. This worked fine for months, but recently there are
periods of an hour or an hour and a half, starting at a fixed time like 12am
or 4am where pdns reports that the backend times out.

For example....

Jan 13 00:01:23 magenta pdns[2150]: Recursive query for remote 83.223.108.22
with internal id 89 was not answered by backend within timeout, reusing id
Jan 13 00:01:51 magenta pdns[2154]: Recursive query for remote 83.223.108.22
with internal id 90 was not answered by backend within timeout, reusing id
Jan 13 00:02:11 magenta pdns[2156]: Recursive query for remote 83.223.126.6
with internal id 6 was not answered by backend within timeout, reusing id
Jan 13 00:02:31 magenta pdns[2150]: Recursive query for remote 83.223.126.6
with internal id 0 was not answered by backend within timeout, reusing id
Jan 13 00:02:41 magenta pdns[2154]: Recursive query for remote 83.223.120.38
with internal id 1 was not answered by backend within timeout, reusing id
Jan 13 00:02:51 magenta pdns[2156]: Recursive query for remote 83.223.120.38
with internal id 2 was not answered by backend within timeout, reusing id
Jan 13 00:03:01 magenta pdns[2150]: Recursive query for remote 83.223.126.6
with internal id 3 was not answered by backend within timeout, reusing id

This continues for about half an hour, and then the logs start showing...

Jan 13 00:38:09 magenta pdns[2144]: Received packet from recursor backend
with id 4 which is a duplicate
Jan 13 00:38:19 magenta pdns[2144]: Received packet from recursor backend
with id 5 which is a duplicate
Jan 13 00:38:29 magenta pdns[2144]: Received packet from recursor backend
with id 7 which is a duplicate
Jan 13 00:38:39 magenta pdns[2144]: Received packet from recursor backend
with id 6 which is a duplicate
Jan 13 00:38:49 magenta pdns[2144]: Received packet from recursor backend
with id 3 which is a duplicate
Jan 13 00:38:59 magenta pdns[2144]: Received packet from recursor backend
with id 4 which is a duplicate
Jan 13 00:39:09 magenta pdns[2144]: Received packet from recursor backend
with id 5 which is a duplicate

And from then on, we get a mixture of the two until it decides to start
working again. Nothing in the bind logging shows anything untoward, so it
looks like for some reason pdns just can't be bothered to forward it to
bind.

The weirdest thing is that it doesn't happen every day, but when it does
it's on the hour for a fixed period of time.

So..... Has anyone seen this before? ;)


Many thanks


Olly Butterfield