We see random timeouts on challenges. For example tohuto .com. Without www it is OK, but www subdomain gets timeout. Everything seems to be the same (HTTP redirects to HTTPS). Do you see some more details about this timeout? It seems to be ipv6 related, but there is no logic that without “www.” it works…
Hi @silver, sorry to hear you're having a random seeming problem with validation. That's always frustrating!
Looking at the most recent failure for www.tohut.com our side started the validation to 2a02:29e8:700:0:1::90 at 2017-12-08 14:26:06.955181+00:00. The result was recorded as a timeout at 2017-12-08 14:26:17.522137+00:00 which is approximately 10s afterwards (the configured single dial timeout). The specific error was "Client.Timeout exceeded while awaiting headers" - that generally indicates the connection was successful but that the HTTP server took too long to respond to the validation request.
What makes you think it's IPv6 related? I thought so initially but the successful validations for tohuto.com without the www. are to the same IPv6 address: 2a02:29e8:700:0:1::90.
It might be useful to try and use tcpdump or tshark to record the validation attempts and your replies. If you can see your webserver replying to the requests in <10s after receiving them I can help debug further.
We still think its ipv6 problem. Can you please show taceroute from your side to 2a02:29e8:770:0:3::26? We see some major packetloss over Cogent. This started 6th december already and we see mod_reqtimeout responses to your requests. We rerouted from our side over another uplink provider, but we can’t see how is your trace coming.
Looks like our reroute helped and we do not see timeouts anymore.
But, while analyzing logs, we see that if domain resolves both to ipv4 and ipv6, then you try only ipv6 (no random nor fallback to ipv4 on timeout).
Maybe you can implement https://en.wikipedia.org/wiki/Happy_Eyeballs