Recently I've installed luci-app-banip on my OpenWrt router and blocked most countries from accessing my services on my network. Not seeing why I would want any of that traffic I also blocked the whole of the ARIN registry, responsible for IP addresses from Canada and the United States.
Edit: Note this is only for inbound traffic. Outbound traffic is allowed no matter the target country.
Fast forward a few weeks and my certbot renewals fail with the following error: Failed to renew certificate enter.domain.here with error: HTTPSConnectionPool(host='acme-v02.api.letsencrypt.org', port=443): Read timed out. (read timeout=45)
Confused af I start looking for solutions and as so often only find useless or completely ridiulous solutions (lowering my MTU to 1300, what? WHY?). Finally I find some enlighted figure that says they recently enabled a blocklist for certain countries and that was the issue for them.
Now I make the connection to my use of banIP, re-allow the USA and my cert renewals start working again. Hooray!
However, there are two things bothering me:
- Why would such a block even interrupt my renewals? I'm using DNS challenges and the ACME servers should only check the DNS entries, not where those entries actually redirect to. The DNS server/root isn't in my home network, so isn't affected by any firewall shenanigans I do here.
- How can I make an exception for the Let's Encrypt ACME servers while blocking the rest of the ARIN IP space?
I see there's the option for ASN selection and external allowlists:
Does anybody have an idea on how to configure this so that Let's Encrypt continues to work without compromising on my network security?
(Edit: And just for clarity, I do not live in the US or anywhere on the American continent.)
are you using external DNS hosting? is it in a (now) blocked country? if so, then your local certbot is unable to update the DNS server records (return traffic from your DNS host is being blocked by your iptables/nftables config).
yeah, that would suggest an https renewal method. had you previously configured web server renewal at all before switching over to DNS? any other suspicious notifications in the logs?
edit: in thinking about this a little more... the renewal has to be initiated by your host, and that is likely done via https (you talk https to the acme server and tell it you want a renewal by DNS). so, if you are blocking the acme servers then the same issue applies - no return traffic.
I've been using DNS challenge for this domain from the start. I'm not sure what you mean by external DNS hosting. The domain is from netcup, the certbot host runs in my local network (as does the HTTP server that the domain points to).
Netcup is a German hosting company, I live in Germany, inbound traffic from Germany is NOT blocked on my router, outbound traffic isn't blocked at all.
not hosting your own DNS server. specifically it sounds like your DNS server is hosted on your domain provider, not your own local network. you have set up certbot to automatically configure your remotely hosted DNS server for the DNS based renewal.
if DNS based recert was working before then it should be working now.
as I said in my edit, you are likely blocking the return https traffic from the US based let's encrypt acme servers - so your initial diagnostic is correct. your local firewall is likely stopping the acme servers from talking back to your local host.
you are right back where you started, asking for info in how to allow-list the acme IP ranges. but at least we may now know why it is not working and you are seeing an https timeout even though you are using DNS based certificate renewals.
edit: typos
AFAICT it is using DNS challenges, unless the cerbot netcup plugin somehow does stuff it shouln't need to do.
agreed. you are using DNS-01 challenges. so the workflow is...
your local certbot machine initiates an https connection to the letsencrypt servers to start the DNS-01 challenge. during this HTTPS dialog, your local certbot is informed of the key material to insert into your DNS records. your local certbot then modifies your netcup DNS server (hosted remotely, not on your local network) with the keying material and the letsencrypt servers verify that the keys are actually there, proving that you control the domain. the letsencrypt serves then issue you the certificate (again, via HTTPS) and your local certbot stores it in your local host.
the issue is most likely stems from the initial HTTPS connection that certbot tries to make to the let's encrypt servers. while your firewall allows this traffic out, it does not allow return traffic back in because of your explicit blocking of US (and perhaps other) based addresses.
even through your are using DNS for your domain autentocation, your local host - the machine running certbot - is unable to initiate the certificate transfer because of the firewall blocking return traffic.
the two external networks (and, therefore IP ranges/subnets/etc) that are important here are the let's encrypt servers and the netcup DNS servers. certbot will have to talk to both of these in order to function.