Failover to other DNS nameservers

[drnic]

From what I was observing (v0.2.X), it seems that if I add consul agent as the first nameserver in /etc/resolv.conf and a hostname isn't resolved by consul, then it doesn't failover to the other nameservers [1] as other DNS nameservers have done for me in the past. Which chunk of consul code is doing this, and how should it fail nicely as a resolv nameserver if it cannot resolve a hostname (nor does its optional recursor chain)?

[1] I know consul also supports a single explicit recursor

[drnic]

I'm assuming that consul is the primary nameserver and it allows 0+ secondary nameservers. It doesn't seem to allow the secondaries a chance to resolve hostnames.

[salehe]

@drnic AFAIK the behavior of linux resolver is to failover to next nameservers defined in /etc/resolv.conf only if the previous one is timed out. Once a nameserver responds to the query (either with failure or success), the resolver process is over.
In your case, if consul agent is available, secondary nameservers are always bypassed.

[armon]

This is odd indeed, since we didn't change the way DNS works at all between 0.2 and 0.3. The only changes were to optionally allow for TTL values or stale reads. Default behavior is unchanged. I agree with @salehe however, that the secondaries are only used if the first DNS server fails to respond. Relying on that behavior for DNS recursion is not the way to go.

[armon]

Closing, but please re-open if an issue.