An ACME validation failure usually shows up as a short line in a cron log at 03:00, and nobody reads it until the certificate expires. This article covers why ACME validation challenges fail across HTTP-01, DNS-01 and TLS-ALPN-01, how to read the error the CA returns, and how to fix the cause before a renewal window closes. If you run Certbot, acme.sh, Caddy, Traefik or cert-manager, the same handful of root causes accounts for most failed renewals.
Fixing these failures matters more now than it did a few years ago. Under CA/Browser Forum ballot SC-081, maximum certificate lifetimes fell to 200 days in March 2026 and will reach 47 days by March 2029. Each step down means more renewals, and every renewal is another chance for validation to fail.
What the CA Actually Checks During Validation
The CA has to confirm that you control the domain. With HTTP-01, it fetches a token from http://yourdomain/.well-known/acme-challenge/ over port 80. DNS-01 looks for a TXT record at _acme-challenge.yourdomain. TLS-ALPN-01 opens a connection on port 443, negotiates the acme-tls/1 protocol and expects a special self-signed certificate.
Many troubleshooting guides leave out one detail. Since 2024, Let’s Encrypt has validated from several network perspectives at once, and most of them have to agree. If your server answers correctly from one region but a firewall or geo-blocking rule drops traffic from another, validation fails even though your own test passed. The full ACME issuance flow explains where this check happens in the order lifecycle.
Reading ACME Challenge Errors Correctly
Most clients print the CA’s problem type, and that type tells you what went wrong. An experienced SRE checks it before touching any configuration.
urn:ietf:params:acme:error:unauthorized means the CA reached something but got the wrong content back. Typical causes are a 404, a stale token or a default vhost answering instead of the right site. connection means the CA could not reach your server, which points to firewalls, closed port 80 or broken IPv6. dns means the CA could not resolve the name or the TXT record. caa means a CAA record forbids this CA from issuing. rateLimited means repeated failures have locked you out for a while. Let’s Encrypt allows 5 failed validations per account, per hostname, per hour.
Common Reasons ACME Validation Challenges Fail
Stale AAAA records are among the most frustrating causes. Let’s Encrypt prefers IPv6 when a AAAA record exists. If someone migrated the site to a new VPS and updated only the A record, the CA may connect to an old IPv6 address that serves nothing. On a dual-stack laptop, curl quietly falls back to IPv4, so the problem never shows up locally.
Proxies and CDNs intercepting the challenge path come next. Cloudflare, Fastly or an AWS ALB may answer /.well-known/acme-challenge/ from a cache, return a WAF block page, or route the request to an origin that doesn’t hold the token. Cloudflare’s proxy also terminates TLS, which breaks TLS-ALPN-01 entirely. CDN configuration problems cause a large share of the renewal tickets that land on platform teams.
DNS-01 propagation and delegation issues are different. Let’s Encrypt queries authoritative nameservers rather than your resolver, so “it works with dig 8.8.8.8” proves nothing. Split-horizon DNS, a CNAME on _acme-challenge pointing to a delegated zone the client can’t write to, or a provider API that returns success before every authoritative server has the record will all produce failures. DNS changes that break certificates often surface first as a failed renewal.
CAA records fail silently until renewal day. A security team adds “0 issue digicert.com” for a new OV certificate, and the next Let’s Encrypt renewal on that zone is refused.
Myth: HTTPS Redirects Break HTTP-01
Many admins assume they must exclude the challenge path from their HTTP-to-HTTPS redirect. That’s wrong. The CA follows up to 10 redirects, to HTTP or HTTPS on ports 80 and 443, and it does not validate the certificate on the HTTPS target. An expired or self-signed certificate on the redirect target is fine.
What does break validation is redirecting to a different port such as 8443, redirecting to a host that doesn’t serve the token, or not having port 80 open at all. Port 80 has to be reachable for HTTP-01 even if the site itself only serves HTTPS.
Step-by-Step Fix for a Failing Renewal
A seasoned DevOps engineer works through failures in a fixed order rather than rerunning the client and hoping:
1. Read the exact problem type and the address the CA tried, which most clients log in verbose mode (certbot -v, acme.sh –debug 2).
2. Test from outside your network, over both IPv4 and IPv6: curl -4 and curl -6 against http://domain/.well-known/acme-challenge/test with a placeholder file.
3. For DNS-01, query each authoritative server directly with dig @ns1.provider.com TXT _acme-challenge.domain, and check CAA with dig CAA domain.
4. Fix the cause, then test against the Let’s Encrypt staging endpoint (certbot –dry-run) so production rate limits stay untouched.
5. Run the real renewal and confirm the new certificate is actually being served, not just written to disk.
Step 5 gets skipped constantly. Certbot renews the file, but nginx keeps serving the old certificate from memory because no deploy hook reloads it. The renewal “succeeded” and the site still expires.
Mistakes That Turn a Small Failure Into an Outage
Trusting exit codes instead of the served certificate. A renewal job can return 0 while the load balancer still holds last quarter’s certificate. Expiry monitoring has to check what clients actually receive over TLS, not what sits in /etc/letsencrypt/live.
Retrying in a loop after failure. A systemd timer that retries every five minutes uses up the 5-per-hour failed validation limit before anyone looks at it. The limit resets, but the logs fill with rateLimited errors that hide the original cause.
Ignoring the failure email until day 1. With 90-day certificates, Certbot renews at 30 days remaining, so there’s a month of slack. With 47-day certificates, renewal typically starts around day 30-33, which leaves roughly two weeks. A failure on a Friday before a holiday weekend can now realistically end in an outage.
There are exceptions. Air-gapped or internal-only hosts can’t pass HTTP-01 at all, and DNS-01 with a scoped API token is the only practical option. Teams in regulated environments that cannot store DNS API credentials on web servers often use CNAME delegation to a dedicated validation zone, such as acme-dns, so a compromised web host can’t rewrite production DNS.
FAQ
How long should I wait for DNS-01 propagation?
Wait until every authoritative nameserver returns the TXT record, not for a fixed time. Most providers need 10-120 seconds, but some managed DNS services take several minutes. acme.sh and cert-manager can poll the authoritative servers before asking the CA to validate.
Why does validation fail when the site loads fine in my browser?
Your browser uses your network path, resolver and usually IPv4. The CA validates from several external locations and prefers IPv6, so geo-blocking, stale AAAA records or a WAF rule can block it while you see nothing wrong.
Can I use HTTP-01 for wildcard certificates?
No. Wildcard certificates require DNS-01. HTTP-01 and TLS-ALPN-01 only validate specific hostnames.
Summary: Make Validation Failures Visible Early
ACME challenges fail for a small set of reasons: unreachable ports, stale DNS records, proxies intercepting the challenge path, CAA restrictions and propagation delays. The problem type in the error message usually narrows it down within minutes. The habit worth building is to verify the served certificate independently of the renewal client and to treat a single failed renewal as an incident while there are still weeks left, not hours.
