Health and Readiness Checks

Liveness answers “is this process running?” Readiness answers “can this process serve real traffic?” Keeping them separate prevents a temporary database outage from restarting a healthy process.

Setup

mkdir zinc-health
cd zinc-health
go mod init example.com/zinc-health
go get github.com/0mjs/zinc
go get github.com/jackc/pgx/v5/stdlib

Application

package main

import (
  "context"
  "database/sql"
  "log"
  "time"

  "github.com/0mjs/zinc"
  _ "github.com/jackc/pgx/v5/stdlib"
)

func main() {
  db, err := sql.Open("pgx", "postgres://localhost/app?sslmode=disable")
  if err != nil {
    log.Fatal(err)
  }
  defer db.Close()

  app := zinc.New()

  app.Get("/live", func(c *zinc.Context) error {
    return c.JSON(zinc.Map{"status": "up"})
  })

  app.Get("/ready", func(c *zinc.Context) error {
    ctx, cancel := context.WithTimeout(c.Context(), 500*time.Millisecond)
    defer cancel()

    if err := db.PingContext(ctx); err != nil {
      return c.Status(zinc.StatusServiceUnavailable).JSON(zinc.Map{
        "status":   "unavailable",
        "database": "down",
      })
    }

    return c.JSON(zinc.Map{
      "status":   "ready",
      "database": "up",
    })
  })

  log.Fatal(app.Listen(":8080"))
}

Swap the PostgreSQL driver and connection string for the database your service already uses; the Zinc handlers are unchanged.

Probe it

curl -i http://localhost:8080/live
curl -i http://localhost:8080/ready

Point an orchestrator’s liveness probe at /live and readiness probe at /ready. Keep readiness checks bounded with short timeouts, and include only dependencies required to serve traffic.