health-check-http

HTTP health check

Probe a service by issuing an HTTP GET. The check passes when the server responds with a 2xx status.

Config

# HTTP health check: probe a service by issuing an HTTP GET.
processes:
  - name: web
    command: /usr/local/bin/web
    args: ["--listen", "127.0.0.1:PORT"]
    on-failure: shutdown

checks:
  web-http:
    http:
      url: http://127.0.0.1:PORT/
    period: 300ms
    threshold: 1

• http.url is fetched every period; a 2xx response is healthy.
• threshold is the number of consecutive failures before the check is marked unhealthy.
• For HTTP over a Unix socket, add socket: /path/to.sock under http: and keep a normal url.
• In a real config, command is your web service and url is its health endpoint (e.g. /healthz). For a quick manual experiment, python3 -m http.server 8080 --bind 127.0.0.1 works as a stand-in service.

Expected behavior

• gopherd starts web and the web-http check.

• Once the server binds, the GET probe returns 200 and the check reports healthy:

  checks:
    web-http             healthy  failures=0
  

• If the endpoint is unreachable or returns non-2xx, the check trends unhealthy once threshold is reached.

Test

Run level. PORT is substituted with an OS-assigned free port and the placeholder command with a small Go HTTP responder built by the test harness (internal/doctest/cmd/httpok) — no external interpreter needed. The test waits for the probe to get a 2xx and asserts the web-http line is healthy.

go test ./documentation/health-check-http/ -v