Running Caddy?

Caddy failing to issue certs or Caddyfile throwing errors? We unstick it.

Caddy is the easiest web server to bring up and the hardest to debug when automatic HTTPS stops working. An ACME challenge that silently fails at 2am, a Caddyfile directive that parses but does the wrong thing, or a reverse proxy to a slow upstream that quietly 502s until the alerting catches it. We read the JSON config and the admin API together until every host serves its intended certificate and every upstream behaves.

  • Since 2015Fixing technology
  • FreeEstimates
  • US teamReal engineers

caddy.log

still failing

What it is

Caddy at a glance

Caddy is a modern open-source web server written in Go and first released in 2015 by Matt Holt. It is best known for automatic HTTPS: configure a hostname and Caddy requests, installs and renews a Let's Encrypt or ZeroSSL certificate without a separate acme-client step.

Features include reverse proxying, HTTP/1, HTTP/2 and HTTP/3, a Caddyfile for human-friendly configuration, a JSON API for programmatic control and an admin endpoint for live reconfiguration. It is used as an edge server, a reverse proxy in front of applications, and a static file server, often instead of Nginx or Apache when HTTPS management was the biggest pain point.

Type
HTTP server with automatic HTTPS
First released
2015
License
Apache 2.0
Maintained by
Matt Holt and the Caddy community
Typically used for
Reverse proxy, static sites, edge TLS termination, automatic HTTPS at small scale

The honest version

Caddy pros and cons

No technology is all upside. This is where Caddy earns its place and where it tends to cause trouble.

Where Caddy is strong

  • HTTPS really is automaticFor almost every hostname you point at Caddy, a certificate shows up and renews itself. Removing the acme-client complexity is a lot more valuable than it sounds until you have done it the hard way.
  • Caddyfile is readableThe human-facing config format is short and legible. A site definition fits in a few lines where Nginx would need twenty and Apache more.
  • Single static binaryWritten in Go and shipped as one binary, Caddy runs the same on bare metal, in a container or on an embedded device without a dependency chain.
  • JSON API for programmatic controlEverything the Caddyfile does is also expressible as JSON, which the admin API accepts and reloads live. Infrastructure tools can drive the server directly.
  • HTTP/3 and modern TLS defaultsThe defaults are current: strong cipher choices, OCSP stapling, HTTP/3 where supported. You do not have to learn the latest best practice to get it.
  • Plugins extend without rebuilding from scratchThe xcaddy tool builds custom Caddy binaries with DNS providers, rate limiters, auth modules and other plugins baked in cleanly.

Where Caddy bites

  • ACME challenges have hidden requirementsHTTP-01 needs port 80 reachable from the internet, TLS-ALPN-01 needs port 443, and DNS-01 needs a provider plugin. A firewall or a misconfigured provider silently blocks renewals.
  • Smaller community than NginxStack Overflow answers, hiring, and tribal knowledge for operations teams all lean Nginx. Teams moving to Caddy need to bring their own expertise.
  • Caddyfile parses but can misleadA directive placed inside the wrong block, or a matcher that applies more broadly than intended, parses cleanly and serves the wrong traffic. The error is in behavior, not in syntax.
  • Observability is basic by defaultLogs are JSON and useful but out of the box, metrics are limited. Prometheus support exists but needs setup to compare with what Nginx users expect.
  • Storage permissions matterCertificates and the account key live in a storage directory. If permissions are wrong or storage is ephemeral, Caddy will keep issuing new certificates and may trip Let's Encrypt's rate limits.
  • Large-scale deployments need careAt thousands of sites on one node, the admin API, storage layer and in-memory config all need tuning. Caddy will do it, but it is not one-click.

Sound familiar?

Common Caddy problems we fix

The Caddy problems that bring people to us, and what you have once each one is fixed.

ACME challenge failed: 403 Forbidden

Automatic HTTPS cannot issue a certificate

Port 80 blocked by a firewall or a load balancer that only forwards 443, a DNS record pointing elsewhere during the challenge, or a Let's Encrypt account that got rate limited.

Certificates issuing and renewing on schedule

ACME: urn:ietf:params:acme:error:rateLimited

Let's Encrypt rate limits are tripped

A storage path that resets on every deploy so Caddy re-issues on every start, or a wildcard request repeated after failures, exhausting the weekly cap.

Storage persistent, issuance within limits

unknown directive in Caddyfile

Caddyfile refuses to load after an edit

A directive that belongs inside a matcher placed at site level, or a feature that needs a plugin built into a custom binary that the current Caddy does not include.

Caddyfile loads and serves cleanly

reverse_proxy: connection refused

Reverse proxy cannot reach the upstream

An upstream process that crashed and was never restarted, a container that bound only to localhost on the wrong interface, or a firewall rule blocking the loopback traffic.

Upstream reachable and serving

no certificates configured for server name

HTTPS requests fall through without a certificate

A hostname in the Caddyfile that Caddy never managed to issue for, often because of a failed challenge that was never retried, or a wildcard hostname without the DNS-01 plugin.

Every served host has a valid certificate

storage: permission denied

Caddy cannot write to its storage directory

A systemd unit running as a user that does not own the data directory, or a container volume mounted read-only.

Storage writable and certificates persist

DNS-01 unsupported by provider plugin

DNS challenge fails on an unsupported provider

A standard Caddy build without the provider plugin for your DNS host, so the DNS-01 challenge is never attempted correctly.

Custom build includes the DNS provider plugin

upstream returned 502 Bad Gateway

Reverse proxy returns intermittent 502s

Upstream processes slow to respond past the reverse-proxy timeout, or health checks that pass just long enough to route traffic to a failing node.

Upstreams healthy, timeouts aligned

What we do

Caddy help, from rescue to upkeep

Reach us for one broken Caddyfile, or let our team run your edge for you.

Rescue
Certificates not renewing, Caddyfile parse errors blocking a reload, or a reverse proxy dropping upstream traffic. We restore service first and explain second.
Caddyfile and JSON config
Human-readable config rewrites that make site definitions short, documented and testable, with the JSON equivalent for programmatic use.
ACME and renewal hardening
Challenge type matched to network setup, storage made persistent across deploys, and alerts on expiry so renewals never silently fail.
Custom Caddy builds
xcaddy builds with DNS providers, rate limiters, auth modules and your own plugins, with CI that keeps them current with upstream releases.
Migration from Nginx or Apache
Moves from Nginx or Apache onto Caddy without changing application behavior, with traffic shadowed before cutover.
Managed Tech
Prometheus metrics, certificate expiry watchers, upstream health dashboards and a US engineer on call when the edge starts dropping requests.

How it works

From first message to fixed

  1. Look Free estimate

    Tell us what's going wrong. We look at it, then tell you plainly what's wrong and what it will cost.

  2. Fix

    We fix the cause, not the symptom. If something is down, getting it back up comes first.

  3. Harden

    Backups, monitoring, updates and access, so the same problem can't quietly come back.

  4. Hand back or stay on

    You get a plain-language write-up. Or we stay on for hosting, upkeep and new work.

Often found together

Running Caddy alongside something else?

We work on the whole stack, so the problem can be wherever it actually is.

Questions

Caddy questions

Something else? Call+1 737 727 7327

Why is Caddy not issuing a certificate for my site?

The three common causes are port 80 blocked at the firewall or load balancer so the HTTP-01 challenge never completes, a DNS record not yet pointing at this server at challenge time, and a Let's Encrypt rate limit reached by prior failed attempts. The logs name which challenge Caddy tried and why it failed; fixing the specific one usually takes minutes.

Should we use Caddy instead of Nginx?

If automatic HTTPS and a short, readable config are the biggest pain, Caddy is often the right move. If you need the massive community, hiring pool and tooling ecosystem that Nginx has, or if you need a specific Nginx module with no Caddy equivalent, stay on Nginx. We have helped teams go in both directions.

How do we avoid tripping Let's Encrypt rate limits with Caddy?

The common cause is storage that resets between deploys, which makes Caddy re-issue every time it starts rather than reusing the existing certificate. Mounting a persistent volume for the storage directory, or pointing Caddy at a shared storage backend, fixes it immediately. For wildcard certs we recommend DNS-01 so retries are cheaper.

Can Caddy sit in front of Node, Rails or Django?

Yes. Caddy's reverse_proxy directive handles Unix sockets, loopback TCP and remote upstreams, and the automatic HTTPS wraps all of it. We wire it up with proper timeouts, health checks and logs so a slow upstream fails gracefully rather than hanging the request.

How do we monitor a Caddy deployment properly?

Enable the Prometheus metrics endpoint, scrape it into your existing monitoring stack, and set alerts on certificate expiry, upstream errors and request latency. We set up the dashboards and the watchdogs so an ACME failure pages someone before it becomes a browser warning.

How much does Caddy help cost?

Estimates are at no charge. Tell us about the deployment, the hostnames involved and what has gone off, and we send a price before any configuration change is made.

Free estimate

Caddy giving you trouble?

Share the deployment, the hostnames and what the logs are saying, and a Hawkins.Tech engineer gets back with an estimate.

  • No charge to find out what it will take
  • If it's urgent, say so

Replies come from a person. By sending, you authorize us to email you back. No marketing.