CertiMonitor

Blog

Certificate lifecycle management for small teams

Jaime Guerrero · · 5 min read

"Certificate lifecycle management" sounds like an enterprise product category, and it is one. But the underlying idea applies to a team with ten certificates just as much as one with ten thousand: every certificate goes through the same stages, and an outage is almost always one stage being nobody's job.

Here are the stages and the tools that exist for each.

1. Know what you have

You can't renew a certificate you don't know about. The forgotten ones cause the outages: a staging host, an old API endpoint, a load balancer someone set up for a migration.

There are three ways to build the list, and you want all three:

  • From your config. Every place that terminates TLS: load balancers, CDNs, ingress controllers, mail servers, VPN appliances.
  • From DNS. Every hostname in your zones that answers on port 443 (or 465, 993, 8443).
  • From Certificate Transparency logs. Every publicly trusted certificate is logged publicly (how that works), so the logs list every certificate ever issued for your domains, including ones issued by a team or vendor you didn't know about. crt.sh is a free search interface: https://crt.sh/?q=%25.yourdomain.com lists certificates for all subdomains.

For each certificate, write down where it's deployed, who owns it and how it renews. The owner is the column that prevents outages.

2. Control who can issue

A CAA record in DNS lists which certificate authorities may issue for your domain. CAs are required to check it before issuing. github.com allows four:

$ dig +short CAA github.com
0 issuewild "sectigo.com"
0 issue "globalsign.com"
0 issue "digicert.com"
0 issue "sectigo.com"
0 issuewild "digicert.com"
0 issuewild "letsencrypt.org"
0 issue "letsencrypt.org"

issue covers normal certificates, and issuewild covers wildcards separately. Two extensions are worth knowing:

  • iodef gives the CA an address to report requests that CAA blocked.
  • accounturi (RFC 8657) restricts issuance to a specific ACME account, so even someone who can pass validation can't get a certificate from a different account at the same CA.

CAA is checked per name, walking up the DNS tree, so a record on yourdomain.com covers every subdomain unless a subdomain has its own.

3. Issue: validation and ACME

Before issuing, the CA confirms you control the domain. That's domain validation (DV), and it's what almost every TLS certificate uses. OV and EV certificates add checks on the organization, but browsers no longer show anything different for them, so they're mostly a procurement requirement now.

ACME (RFC 8555) is the protocol that automates issuance. Every ACME CA publishes a directory. Here is Let's Encrypt's, slightly trimmed:

$ curl -s https://acme-v02.api.letsencrypt.org/directory
{
  "keyChange": "https://acme-v02.api.letsencrypt.org/acme/key-change",
  "meta": {
    "caaIdentities": ["letsencrypt.org"],
    "profiles": {
      "classic": "https://letsencrypt.org/docs/profiles#classic",
      "shortlived": "https://letsencrypt.org/docs/profiles#shortlived",
      "tlsserver": "https://letsencrypt.org/docs/profiles#tlsserver"
    },
    "termsOfService": "https://letsencrypt.org/documents/LE-SA-v1.8-July-06-2026.pdf",
    "website": "https://letsencrypt.org"
  },
  "newAccount": "https://acme-v02.api.letsencrypt.org/acme/new-acct",
  "newNonce": "https://acme-v02.api.letsencrypt.org/acme/new-nonce",
  "newOrder": "https://acme-v02.api.letsencrypt.org/acme/new-order",
  "renewalInfo": "https://acme-v02.api.letsencrypt.org/acme/renewal-info",
  "revokeCert": "https://acme-v02.api.letsencrypt.org/acme/revoke-cert"
}

An issuance goes like this:

  1. The client registers an account, identified by its own key pair (not the certificate's key).
  2. It creates an order for one or more names (newOrder).
  3. For each name the CA returns an authorization with challenges to prove control.
  4. The client completes one challenge per name.
  5. It finalizes the order by sending a CSR containing the certificate's public key, and downloads the certificate and chain.

The challenge type decides what your automation needs access to:

ChallengeHow it proves controlNeedsWildcards
http-01Serves a token at http://name/.well-known/acme-challenge/Port 80 reachable from the internetNo
dns-01Publishes a token in a TXT record at _acme-challenge.nameAn API key for your DNS providerYes
tls-alpn-01Answers a special TLS handshake on port 443Control of the TLS listenerNo

dns-01 is the most flexible and the riskiest to operate: the automation holds a key that can change your DNS. Scope that key to TXT records under _acme-challenge if your provider allows it.

The profiles entry in the directory is newer. Let's Encrypt offers classic (90-day certificates, the default), tlsserver (45 days) and shortlived (about 6 days). Short-lived certificates expire before revocation would even matter, which is the direction the industry is going.

4. Deploy

Issuance isn't the end. The new certificate has to reach every place that serves it, and the software has to load it. The most common renewal failure isn't a failed renewal: it's a successful one, sitting on disk, while the server keeps the old certificate in memory. Every ACME client has a hook for this (certbot's --deploy-hook, for example). Use it to reload the service, and test it once by hand.

Watch the private key too. Generate it where it's used, never email it, and let it rotate: most ACME clients create a new key on every renewal unless told to reuse it, which limits the damage from a key leak.

5. Renew

The traditional rule is to renew when about a third of the lifetime is left: day 60 of a 90-day certificate. That leaves time for a failed renewal to be retried and noticed. As lifetimes shrink (the schedule), the margin in days shrinks with them.

ARI (ACME Renewal Information, RFC 9773, published in 2025) replaces the fixed rule. The client asks the renewalInfo endpoint when to renew each certificate, and the CA answers with a suggested window. Normally that spreads renewals out to avoid load spikes. When a CA has to revoke certificates early, it moves the window forward, and ARI-aware clients replace their certificates before the revocation hits. If your ACME client supports ARI, turn it on.

6. Revoke and retire

Revocation is for when a certificate shouldn't be trusted before its expiry: a leaked key, a domain you no longer control, or a CA mistake. The CA/Browser Forum's rules give CAs 24 hours to revoke in the most serious cases, such as a key compromise, and 5 days for most others. That's why CA incidents occasionally force thousands of customers to replace certificates within days, and why automation (plus ARI) is the only realistic way to handle it.

Retiring is the step most teams skip. When a service is shut down, remove its DNS records and stop renewing its certificate, so it doesn't keep showing up in inventories and alerts, and so the hostname doesn't sit around pointing at an IP someone else may get later.

7. Watch it from the outside

Every stage above has a way of failing quietly: the DNS API key expires, the deploy hook stops reloading, the CAA record blocks your new CA, a team issues a certificate you never hear about. The only check that catches all of them is looking at what each host actually serves, on a schedule, with alerts going to whoever owns that certificate. Monitoring doesn't replace automation, but it's how you find out automation stopped.

The free check shows what a host is serving right now, including the chain and expiry date:

Sources