Nginx returns 502 after I redeploy one container

Question

Hi, I run Nginx and my app in Docker Compose on a Droplet. When I redeploy only the app container (docker compose up -d app), Nginx starts returning 502 until I restart Nginx too. The app is healthy and I can reach it directly on its port. Why does Nginx not find it again, and is there a way to fix this without restarting Nginx every time?

Bobby · Answer

Hi there, Nginx resolves the upstream hostname only once, when it starts. When you recreate the app container, it can get a new internal IP, but Nginx keeps using the old one. That is why you get 502 until you restart or reload it. Quick fix is to reload Nginx after the deploy: docker compose exec nginx nginx -s reload A better fix is to make Nginx resolve the name again at request time. Use Docker’s built in DNS and put the upstream in a variable: server { listen 80; resolver 127.0.0.11 valid=10s; location / { set $upstream http://app:3000; proxy_pass $upstream; } } The resolver line points to Docker’s DNS, and using a variable in proxy_pass forces Nginx to look up the name again every few seconds. One thing to know: when you use a variable, proxy_pass does not cut the URI part for you, so keep your paths simple or handle them with rewrite. Also add a healthcheck to the app and use depends_on with condition: service_healthy, so Nginx does not start before the app is ready. If you only ever have one app container, the reload step in your deploy script is the simplest option. If you want to practice Docker networking and container commands without breaking your own server, this browser based Docker terminal simulator is a safe place to try things out.

alexdo · Answer

Hi there, To add to Bobby’s answer, if your Nginx image is 1.27.3 or newer, you can keep a normal upstream block and still have Nginx pick up the new container IP. Before 1.27.3 the resolve parameter on upstream servers was only available in NGINX Plus. It is now part of open source Nginx, and the official nginx Docker images ship current versions. upstream app_backend { zone app_backend 64k; resolver 127.0.0.11 valid=5s; server app:3000 resolve; } server { listen 80; location / { proxy_pass http://app_backend; } } The zone line is required, because resolve needs the upstream group in shared memory. 127.0.0.11 is Docker’s embedded DNS, as in Bobby’s example. Compared to the variable approach, proxy_pass keeps its normal URI handling, so you don’t need the rewrite workaround, and you can still use upstream features like keepalive. You can check your version with: docker compose exec nginx nginx -v Two small things that can still cause a few 502s during a redeploy, even with this in place: docker compose up -d app stops the old container before the new one is ready, so there is a short window with nothing to proxy to. If you want to avoid that, you need two app instances behind the upstream, or a tool like docker-rollout that starts the new container before stopping the old one. The healthcheck with depends_on: condition: service_healthy only controls startup order when Nginx itself is started. It does not hold back traffic when you recreate just the app container, so it won’t cover the redeploy window above. Regards

Become a contributor for community

DigitalOcean Documentation

Resources for startups and AI-native businesses

The developer cloud

Start building today

Report this

Nginx returns 502 after I redeploy one container

Related questions

How do I get Flash and Nginx to Communicate with each other?

How to tell if your reverse proxy is buffering your token stream (and why I had to rent a Droplet to find out)

How can I improve the speed of an e-commerce website?

How to optimize Nginx reverse proxy caching for media streaming apps?

How to configure Nginx reverse proxy on Ubuntu

How to fix in nginx security config

How to change existing certbot/nginx to wildcard domain?

How do I get Ruby with Hanami to serve dynamic files via Falcon?

Unable to access the URL after deploying Discourse

AstroJS project hosted on DigitalOcean droplet with Nginx – CSS loading as text/html

Setting up a reverse proxy on DigitalOcean App Platform to route main domain and subdomains

using shiny-server with nginx

Report this

Nginx returns 502 after I redeploy one container

Related questions

How do I get Flash and Nginx to Communicate with each other?

How to tell if your reverse proxy is buffering your token stream (and why I had to rent a Droplet to find out)

How can I improve the speed of an e-commerce website?

How to optimize Nginx reverse proxy caching for media streaming apps?

How to configure Nginx reverse proxy on Ubuntu

How to fix in nginx security config

How to change existing certbot/nginx to wildcard domain?

How do I get Ruby with Hanami to serve dynamic files via Falcon?

Unable to access the URL after deploying Discourse

AstroJS project hosted on DigitalOcean droplet with Nginx – CSS loading as text/html

Setting up a reverse proxy on DigitalOcean App Platform to route main domain and subdomains

using shiny-server with nginx