1. The problem I’m having:
My caddy reload stopped working a few days ago.
Please note that my caddy.json configuration file is quite big, more than 10MB, and contains a lot of different routes.
The issue only happens when I reload. When I initially run caddy with this configuration file, caddy works as expected.
Little bit more about my setup
I have the initial caddy.json that loads the real configuration file that’s hosted remotely.
For the sake of this post, let’s call these two caddy config files initial and final config file.
Whenever I need to handle a new domain, I generate the new final config, and then reload caddy with caddy reload --config caddy.json --force. This used to work like charm and caddy would start handling the new domain immediately upon reload.
I use grace_period set to 10 seconds.
2. Error messages and/or full log output:
Not exactly an error message, but this is what I found while debugging so maybe it will help you help me 
When I fetch the config file via GET localhost:2019/config, I used to get the final configuration file. But few days ago, all the GET /config requests started returning the initial config file (the one that should load the real configuration file).
I feel that even though caddy reload --config caddy.json --force returned with success, because it’s an asynchronous call, it hangs and never completes. And my only solution is to keep restarting my server so caddy picks up the new configuration file, which is not ideal as you can imagine 
3. Caddy version:
2.10.0
4. How I installed and ran Caddy:
a. System environment:
Docker (Linux)
b. Command:
caddy run --config /etc/caddy/caddy.json
d. My complete Caddy config:
NOTE: I can share my production caddy config file via DM with @matt, @francislavoie or someone else from the Caddy team.
caddy.json (initial
{
"admin": {
"config": {
"load": {
"method": "GET",
"module": "http",
"adapter": "json",
"url": "https://gist.githubusercontent.com/popcorn/b569e72d2223a0ae96fe97f6f88efe93/raw/53c092a5d698e927df68946b2e849f2bbba11c9c/caddy.json"
}
}
}
}
caddy.json (final config file that gets loaded)
This gist is public so you can fully test it.
{
"apps": {
"http": {
"grace_period": "10s",
"servers": {
"tls_terminator": {
"listen": [
":443"
],
"routes": [
{
"match": [
{
"host": [
"localhost"
]
}
],
"handle": [
{
"handler": "static_response",
"body": "Hello, world!",
"status_code": 200
}
]
}
],
"logs": {}
}
}
}
},
"admin": {
"identity": {
"issuers": [
{
"module": "acme",
"email": "drago@gmail.com+latest"
}
]
}
},
"storage": {
"module": "file_system",
"root": "~/caddy-dev/data"
}
}
Here’s more information on how I run this.
This is my Dockerfile:
FROM caddy:2.10.0-builder AS builder
RUN xcaddy build \
--with github.com/ss098/certmagic-s3 \
--with github.com/caddyserver/replace-response \
--with github.com/caddy-dns/cloudflare \
--with github.com/zhangjiayin/caddy-geoip2 \
--with github.com/mholt/caddy-ratelimit \
--with github.com/pberkel/caddy-storage-redis \
--with github.com/caddyserver/cache-handler \
--with github.com/darkweak/storages/go-redis/caddy
FROM caddy:2.10.0
# Install bash and curl
RUN apk update && apk add bash curl
COPY --from=builder /usr/bin/caddy /usr/bin/caddy
# Copy the Caddy configuration file to the container
COPY config/caddy-dev.json /etc/caddy/caddy-dev.json
# Copy the start script and make it executable
COPY /bin/start.sh /usr/local/bin/start.sh
RUN chmod +x /usr/local/bin/start.sh
ENTRYPOINT ["/usr/local/bin/start.sh"]
My start.sh script basically only contains:
#!/bin/bash
# Start Caddy in the background
caddy run --config /etc/caddy/caddy.json &
# Wait for all background processes to finish
wait
I build this Docker image with:
docker build -t tls_reverse_proxy_dev -f Dockerfile .
Then I run it with:
docker run -d --name tls_reverse_proxy_dev \
-p 80:80 -p 443:443 \
-v ~/code/tls-reverse-proxy/tls_reverse_proxy_dev/data:/root/caddy-dev/data \
-v ~/code/tls-reverse-proxy/tls_reverse_proxy_dev/log:/var/log/caddy-dev \
-v ~/code/tls-reverse-proxy/tls_reverse_proxy_dev/config/caddy-dev.json:/etc/caddy/caddy-dev.json \
tls_reverse_proxy_dev
Ok, so with this setup, GET /config returns the final configuration file as expected:
wget -qO- 127.0.0.1:2019/config | jq .
{
"admin": {
"identity": {
"issuers": [
{
"email": "drago@gmail.com+latest",
"module": "acme"
}
]
}
},
"apps": {
"http": {
"grace_period": "10s",
"servers": {
"tls_terminator": {
"listen": [
":443"
],
"logs": {},
"routes": [
{
"handle": [
{
"body": "Hello, world!",
"handler": "static_response",
"status_code": 200
}
],
"match": [
{
"host": [
"localhost"
]
}
]
}
]
}
}
}
},
"storage": {
"module": "file_system",
"root": "~/caddy-dev/data"
}
}
But in my production environment, this same command now returns the initial caddy config file:
wget -qO- 127.0.0.1:2019/config | jq .
{
"admin": {
"config": {
"load": {
"method": "GET",
"module": "http",
"adapter": "json",
"url": "https://gist.githubusercontent.com/popcorn/b569e72d2223a0ae96fe97f6f88efe93/raw/53c092a5d698e927df68946b2e849f2bbba11c9c/caddy.json"
}
}
}
}
Does anyone have any idea why caddy reload would hang and never actually execute?