[Dnsmasq-discuss] Dealing with a unstable/slow upstream DNS

Simon Kelley simon at thekelleys.org.uk
Fri Oct 4 15:55:22 BST 2013


On 04/10/13 12:23, Joachim Berdal Haga wrote:
> On 4 October 2013 12:32, Simon Kelley<simon at thekelleys.org.uk>  wrote:
>> On 04/10/13 08:56, Joachim Berdal Haga wrote:
>
>>> Is it possible to instruct dnsmasq to not forward AAAA requests
>>> when the A record is deduced from /etc/hosts (or vice versa)?
>>
>> No, but you're not the first person to ask for this feature. I've been
>> hesitant to add it because it introduces a form of pollution into the DNS
>> that will make deploying IPv6 more difficult.
>
>
> Indeed. One could argue that /etc/hosts already overrides upstream
> DNS, so the pollution is already there. And it makes a certain sense
> to let /etc/hosts be authoritative for a given host name. But the IPv6
> transition is hard enough already. I certainly won't press for this.
>
>
>>> Is it possible to instruct dnsmasq to refresh its
>>> cache while replying immediately with the already cached record? Either
>>> optimistically (when the entry will soon expire) or lazily (when the entry
>>> has just expired).
>>
>> No again. I once saw a patch to dnsmasq that did this for a number of
>> configured domains. You can't really do it for all domains, since it will
>> make the cache grow without bound - one a domain is in you can't ever lose
>> it again.
>
> My thought was not to do this automatically, but only when the record
> is requested by a client. If this request happens within a
> configurable time window (say the last minute before expiry), the
> record is refreshed asynchronously. This would at most double the
> cache size (if every record is refreshed once but not used again
> afterwards).

Indeed, that would be simple to implement, I think. Should revisit this 
after 2.67 is released.


Cheers,

Simon.

>
> Regards,
> Joachim.
>




More information about the Dnsmasq-discuss mailing list