Operator: fix bug where /-/ready and /-/healthy always returned 404 (#…

…1252)

* operator: fix bug where /-/ready and /-/healthy always returned 404

controller-runtime must have at least one ready/healthy check for the endpoints to exist

* fix lint error, use healthz.Ping

CHANGELOG.md

@@ -40,6 +40,9 @@ for detailed information.
 
 - [BUGFIX] Fix panic when using 'stdout' in automatic logging (@mapno)
 
+- [BUGFIX] Grafana Agent Operator: The /-/ready and /-/healthy endpoints will
+  no longer always return 404 (@rfratto).
+
 - [DEPRECATION] The node_exporter integration's `netdev_device_whitelist` field
   is deprecated in favor of `netdev_device_include`. Support for the old field
   name will be removed in a future version. (@rfratto)

cmd/agent-operator/main.go

@@ -12,6 +12,7 @@ import (
 	"github.com/grafana/agent/pkg/operator/logutil"
 	"github.com/prometheus/common/version"
 	controller "sigs.k8s.io/controller-runtime"
+	"sigs.k8s.io/controller-runtime/pkg/healthz"
 
 	// Needed for clients.
 	_ "k8s.io/client-go/plugin/pkg/client/auth"
@@ -33,6 +34,13 @@ func main() {
 		os.Exit(1)
 	}
 
+	if err := m.AddReadyzCheck("running", healthz.Ping); err != nil {
+		level.Warn(logger).Log("msg", "failed to set up 'running' readyz check", "err", err)
+	}
+	if err := m.AddHealthzCheck("running", healthz.Ping); err != nil {
+		level.Warn(logger).Log("msg", "failed to set up 'running' healthz check", "err", err)
+	}
+
 	if err := operator.New(logger, cfg, m); err != nil {
 		level.Error(logger).Log("msg", "unable to create operator", "err", err)
 		os.Exit(1)