I stopped using certbot when they made it nothing but a snap.
Lego, acme.sh, and tools like caddy, traefik, cert-manager can easily replaced it afaiac.
Those look some really cool options! I might end up using one of these for my environment.
Can confirm that acme.sh is a great option. Way better support of many DNS APIs than Certbot, including easier setup of wildcard certificates. Personally moved to this when Certbot's ability to do RFC2136 (dynamic DNS method that many DNS servers support) was seriously lacking, and never looked back.
Iโve used traefik for 7 years at this point and the only time I had to think about certificates was when I blocked my servers running traefik from making DNS calls needed for the cert generation.
Iโve got 6 domains now all with certs managed by traefik. Highly recommend checking it out, especially if youโre running most things in docker.
Traefik is a godsent. Just build your services with compose, add a few labels and most services work directly. If a service needs additional headers or whatever, that's just more labels but traefik takes care of it all.
Especially the certificate function makes the whole deal so much more comfortable
No. Not everyone uses traefik or caddy
Yeah man, that's the point of the article. Its asking the question "should everyone who isnt using them already move to them". Its not saying everyone already does.
cert-manager works really well.
I started out using certbot, but once i needed a reverse proxy i found caddy. I was confused at first at how to set up the certificates for caddy, but it told me it would just work and my sites have the https and the little lock, so i guess it is just magic!
Have found caddy to be generally easy. I think first starting with it took a bit more to figure out, but it does work well
I use certbot on only a single one of my oldest projects that has been going for almost a decade.
For everything else I use acme.sh because it works so well and integrates with a ton of DNS providers. The one time I had an issue, it was already fixed in a PR, so I just checked out that fixed version and used it for renewals until it was merged in.
Using Caddy for couple of years already at home, yet using certbot at job, because of requirements to use nginx as balancer.
Certbot is so problematic we still pay for most of our certificates because itโs more reliable.
Iโm not sure if Caddy/Traefik is the answer but itโs clear the work should be handed over to a team with a proper focus on reliability.
Can you elaborate on this reliability issue?
Certbot is supposed to automatically renew certificates. It doesn't do that reliably in my experience.
We use it on non-critical systems and every few months I need to go in and fix things... that never happens with traditional certificates - those are setup and forget.
As for the exact problems, I don't think we've ever had the same problem twice. It's always a once off thing but it's still an hour of wasted time each and every time. If it happened on a proper production system it'd be a lot more than an hour, since whatever change is made would need a full gamut of testing / reporting / etc.
I stopped using certbot when they made it nothing but a snap.
Lego, acme.sh, and tools like caddy, traefik, cert-manager can easily replaced it afaiac.
Those look some really cool options! I might end up using one of these for my environment.
Can confirm that acme.sh is a great option. Way better support of many DNS APIs than Certbot, including easier setup of wildcard certificates. Personally moved to this when Certbot's ability to do RFC2136 (dynamic DNS method that many DNS servers support) was seriously lacking, and never looked back.
Iโve used traefik for 7 years at this point and the only time I had to think about certificates was when I blocked my servers running traefik from making DNS calls needed for the cert generation.
Iโve got 6 domains now all with certs managed by traefik. Highly recommend checking it out, especially if youโre running most things in docker.
Traefik is a godsent. Just build your services with compose, add a few labels and most services work directly. If a service needs additional headers or whatever, that's just more labels but traefik takes care of it all. Especially the certificate function makes the whole deal so much more comfortable
No. Not everyone uses traefik or caddy
Yeah man, that's the point of the article. Its asking the question "should everyone who isnt using them already move to them". Its not saying everyone already does.
cert-manager works really well.
I started out using certbot, but once i needed a reverse proxy i found caddy. I was confused at first at how to set up the certificates for caddy, but it told me it would just work and my sites have the https and the little lock, so i guess it is just magic!
Have found caddy to be generally easy. I think first starting with it took a bit more to figure out, but it does work well
I use certbot on only a single one of my oldest projects that has been going for almost a decade.
For everything else I use acme.sh because it works so well and integrates with a ton of DNS providers. The one time I had an issue, it was already fixed in a PR, so I just checked out that fixed version and used it for renewals until it was merged in.
Using Caddy for couple of years already at home, yet using certbot at job, because of requirements to use nginx as balancer.
Certbot is so problematic we still pay for most of our certificates because itโs more reliable.
Iโm not sure if Caddy/Traefik is the answer but itโs clear the work should be handed over to a team with a proper focus on reliability.
Can you elaborate on this reliability issue?
Certbot is supposed to automatically renew certificates. It doesn't do that reliably in my experience.
We use it on non-critical systems and every few months I need to go in and fix things... that never happens with traditional certificates - those are setup and forget.
As for the exact problems, I don't think we've ever had the same problem twice. It's always a once off thing but it's still an hour of wasted time each and every time. If it happened on a proper production system it'd be a lot more than an hour, since whatever change is made would need a full gamut of testing / reporting / etc.