Merge pull request #18391 from aweiteka/prometheus-reload

Automatic merge from submit-queue (batch tested with PRs 18780, 18802, 18391, 18832, 18808).

Prometheus config reload livenessprobe

Use case: config updated via configmap. When the change lands in the container (~60 seconds delay) the config directory hash changes and the process is killed. This does not kill the pod but results in a silent reload of config with a corresponding metric timestamp, `prometheus_config_last_reload_success_timestamp_seconds`.

Signed-off-by: Aaron Weitekamp <[email protected]>

Author: openshift-merge-robot

examples/prometheus/prometheus.yaml

@@ -283,6 +283,20 @@ objects:
           - --web.listen-address=localhost:9090
           image: ${IMAGE_PROMETHEUS}
           imagePullPolicy: IfNotPresent
+          livenessProbe:
+            exec:
+              command:
+              - /bin/bash
+              - -c
+              - |-
+                set -euo pipefail;
+                touch /tmp/prometheusconfig.hash;
+                if [[ $(find /etc/prometheus -type f | sort | xargs md5sum | md5sum) != $(cat /tmp/prometheusconfig.hash) ]]; then
+                  find /etc/prometheus -type f | sort | xargs md5sum | md5sum > /tmp/prometheusconfig.hash;
+                  kill -HUP 1;
+                fi
+            initialDelaySeconds: 60
+            periodSeconds: 60
           volumeMounts:
           - mountPath: /etc/prometheus
             name: prometheus-config