martin f krafft wrote:
Package: dnsmasq
Severity: normal
Version: 2.22-2

I am getting the following messages:

  Jul  9 16:14:05 seamus dnsmasq[23117]: forwarding table overflow:
  check for server loops.

Yet my configuration has no loops:

  no-resolv
  server=213.203.238.4
  interface=lo
  bind-interface

and /etc/resolv.conf is:

  nameserver 127.0.0.1
  nameserver 213.203.238.4

The problem is that dnsmasq is started with the -r switch:

  /usr/sbin/dnsmasq -r /var/run/dnsmasq/resolv.conf

so it reads a resolv.conf file (updated by resolvconf) even if
no-resolv is specified in the configuration.


I'm not sure that I agree with your analysis of the problem.

. That -r flag means that /etc/resolv.conf is irrelevant, /var/run/dnsmasq/resolv.conf would be read instead.

. The no-resolv flag does override the -r flag, so no resolv file is being read.

. Even if it were, dnsmasq throws away servers with addresses it's listening on so the 127.0.0.1 entry isn't a problem. To get a server loop, you should need at least two servers forwarding to each other.

The forwarding table overflow message is generated when there are at more than 150 unanswered queries less than 20 seconds old. It's easy to get this with a loop, where the same query bounces around forwever, but it's possible to get it in other ways too. A high query load into dnsmasq and a forwarding to a slow or flaky server will do it: the favourite way is running one of the applications which do bulk-lookups of IP addresses to fill in webserver logs.

Can you give a few more details on the exect problem: is it intermittent for not? What sort of load are you using? What does dnsmasq log about the servers it's using at startup?

It probably worth logging a warning when both resolv-file (-r) and no-resolv are using, since these do conflict. I'll add that to the next release.

Cheers,
Simon.




--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Reply via email to