Proxy Reminder: Race Management Upstream Resilience

Problem

mnmsoft_proxy currently fails hard at nginx startup if the upstream hostname race-management-web is not resolvable at config-parse time.

Observed failure:

host not found in upstream "race-management-web" in /etc/nginx/conf.d/default.conf

This caused the proxy container to restart in a loop whenever the race-management-web container was down or not yet attached to the shared Docker network.

What Was Already Fixed

In /opt/docker/apps/race-management/docker-compose.yml:

db now uses restart: unless-stopped
web now uses restart: unless-stopped

That reduces the chance of the backend staying down, but the proxy should still be made tolerant of temporary upstream absence.

What Needs To Be Done In Proxy

Relevant proxy files:

/opt/docker/apps/proxy/generated/default.conf
possibly /opt/docker/apps/proxy/nginx/nginx.conf

Goal

Avoid nginx resolving race-management-web only once at startup. Instead, make nginx resolve through Docker DNS at request time or with a short-lived resolver cache.

Recommended Approach

Use Docker's internal DNS resolver and proxy through a variable.

Typical pattern:

resolver 127.0.0.11 valid=10s;

location / {
    set $race_management_upstream http://race-management-web:8000;
    proxy_pass $race_management_upstream;
}

Why

With a static upstream like:

proxy_pass http://race-management-web:8000;

nginx may try to resolve the hostname during startup/config load. If the container is absent at that exact moment, nginx exits and the proxy restarts.

Using resolver 127.0.0.11 plus a variable makes nginx rely on Docker DNS more safely at runtime.

Also Clean Up

The proxy logs also show this warning:

the "listen ... http2" directive is deprecated

If present, replace patterns like:

listen 443 ssl http2;

with the newer form appropriate for the installed nginx version.

After Editing

Reload/recreate the proxy stack and verify:

cd /opt/docker/apps/proxy
docker compose up -d
docker logs --tail 100 mnmsoft_proxy

Expected result:

proxy stays up even if race-management-web is briefly unavailable during startup
no more host not found in upstream "race-management-web" fatal error

2.3 KiB Raw Blame History