fix(docker): probe local ClickHouse health over IPv4

Merged
IZ
23d ago
+1-1

Summary

The dev docker-compose.yaml healthcheck ran wget http://localhost:8123/ping. Inside the container localhost resolves to [::1], but ClickHouse listens on IPv4, so the probe is refused every time and the container reports unhealthy forever (84,642 consecutive failures on one machine) regardless of whether the server is fine. That hides real outages: a ClickHouse that is actually hung looks identical to a healthy one.

docker-compose.selfhost.yml already probes 127.0.0.1 since #804; this applies the same fix to the dev compose file.

Test plan

  • Inside the running dev container: wget --spider http://127.0.0.1:8123/ping succeeds; the localhost form is refused
  • docker compose -f docker-compose.yaml config -q

Summary by cubic

Fixes the dev ClickHouse healthcheck so it probes IPv4 instead of localhost, which resolved to [::1] inside the container and never matched ClickHouse's IPv4 listener. The container now correctly reports healthy; docker-compose.selfhost.yml already used 127.0.0.1.