Some of you have noticed that proton has been acting up the last week or so. Specifically, anything that caused a certain type of DNS lookup would fail. People were also having a hard time establishing outbound connections of any type.

begin boring technical explanation

It took quite a bit of digging on our part, but we determined that the problem wasn’t with the DNS resolution at all. There was an invalid default route listed in the routing table. Unfortunately, this default route was only getting used at certain times. If you or the application you’re using explicitly selected a valid interface to use for outbound communication, everything would work fine.

But if the routing was left up entirely to the system, it would default to the first default route listed (which wasn’t valid). So doing a ping from the command line to an external IP would fail, while doing a ping -I eth0 would succeed.

Our initial thought that it was a DNS issue was actually related to this. Local DNS lookups would work just fine, while recursive DNS lookups would fail. It turns out that the static route for the local subnet was allowing the machine to communicate with our local DNS servers just fine. However, when those DNS servers returned a non-authoritative result, the machine would attempt to do a recursive lookup to get an authoritative result, and would then fail because it failed to contact the external name server.

Overall, an annoying problem, but fascinating none the less. Interesting to note the variety of things that can be broken in such different ways by one simple invalid route.

end boring technical explanation

Fortunately, this is all resolved at this point. If anyone has any future problems please contact support and we’ll get you taken care of.