Docs/feature: outbound HTTPS from container fails (Connection reset) – document or support network_mode: host for a service

[mikelito859]

Summary

When deploying apps that download files from external HTTPS URLs from a worker/background process (e.g. Sidekiq downloading WhatsApp media from Meta’s lookaside.fbsbx.com), the download can fail with Connection reset by peer when running inside the default Docker bridge network, while the same request from the host succeeds. This happens on some VPS providers (e.g. Hostinger KVM) and is a known Docker/network behaviour (NAT, IPv6, routing), not an application bug.

Workaround that works: run the affected service (e.g. the worker) with network_mode: host. That requires:

• Exposing Redis/DB on host ports (e.g. 6389:6379, 5433:5432)
• Setting that service’s env vars to use 127.0.0.1 and those ports
• Removing networks: for that service (incompatible with host mode)

Right now this means manually editing the generated docker-compose.yml in the server. Those edits are lost on next deploy because Coolify regenerates the compose.

Suggestion

1. Documentation: Add a short note in the docs (e.g. “Troubleshooting” or “Docker Compose”) explaining that for apps needing stable outbound HTTPS to certain CDNs from a container, bridge network can fail on some hosts, and that using network_mode: host for the affected service (with Redis/DB on host ports) is a valid workaround.

2. Optional – UI: For Docker Compose deployments, consider a per-service option like “Network mode: host” so users can enable it without editing the file, and Coolify can regenerate the compose while preserving that setting (and optionally the needed port mappings / env overrides).

This would help anyone deploying Chatwoot (WhatsApp media), or any similar app that downloads from external URLs inside a worker container.

Context

• Reported in: Chatwoot side: WhatsApp Cloud media download fails in Docker (Connection reset by peer) – workaround: host network for Sidekiq chatwoot/chatwoot#13612
• Observed: Coolify-deployed Chatwoot (Docker Compose), Sidekiq in bridge network; media download fails. Same stack with Sidekiq on network_mode: host (Redis/Postgres on host ports) works.

[Cinzya]

Do you have any evidence supporting your claims?
E.g. tcpdump or strace outputs that would show the TCP reset you are talking about?

Can you also share some more information about your Chatwoots network configuration? (Is connect to Predefined network enabled, did you manually assign other networks on top of it)

I couldn't find much information around cases like yours online. It was often container to container connections not outbound. If there was a outbound problem it seemed to stem from ISP or misconfigured networks.