Marc Pignat wrote: > Hi Simon! > > Thanks you for the explanation! > > I have re-done the test, and it runs almost as you said. > The request is also done in parallel on every server when the ttl is gone > (before the 30 seconds). > My ISP has set a 10 seconds ttl to their hosts...
If the TTL had not expired, dnsmasq would answer from the cache and not forward the query at all. There's one further case that can cause dnsmasq to forward to all servers - if the original requestor re-tries a query. That might be the case here. > > All queries are done with the same source port and dnsmasq seems to close the > port as soon as the first server has replied, generating ICMP > "destination unreachable" messages, is this the intended behavior? I'm not aware of anywhere where it causes problems. The alternative id to keep ports around, but that might cause problems, since ports are a scarce resource and dnsmasq uses a new one for each query. Have you seen problems because of this behaviour? Cheers, Simon. > > > Here is the tcpdump attached, just in case! -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
