Monitor a cluster

A cluster’s page has three tabs for keeping an eye on it: Monitoring shows graphs for each master and worker node, Alerts tells you when a node crosses a threshold, and Activity Timeline lists what was done on the cluster and whether it worked.

Before you begin

This needs a cluster that reads Running. While it reads Creating or Upgrading Master, only its Cluster Details tab works.

Watch a node’s graphs

1. In MyAccount, under Compute, click Kubernetes, click your cluster’s name, and open the Monitoring tab.
2. Pick a node from the list.
3. Choose the period in Showing data:.

The tab shows the node’s graphs of RAM, Storage, CPU Load, CPU Utilization, Disk Read Operation, Disk Write Operation, and Network Traffic over that period. Monitoring and alerts carry no charge of their own.

Set an alert on a node

A new cluster’s nodes come with alerts already set: Trigger : load for average load and Trigger : memory for consumed memory, each above 85%. To see a node’s alerts, open the Alerts tab and pick the node. Each alert is listed with its Alert Type and Alert Condition. To turn one off, pause or delete it there.

To add another alert:

1. On the Alerts tab, pick a node, and click Create Alert. The monitoring alert form opens.
2. Choose a trigger: Processor Load, Consumed Memory, or Free Disk Space.
3. Choose an operator: Greater than, Less than, Equal to, or Not equal to. For load and memory, Greater than alerts you as use climbs; for free disk space, Less than alerts you as the disk fills.
4. Enter the threshold, choose a severity, and choose the alert group to notify. To make an alert group, see Set up alerts on a node.
5. Create the alert.

To view and edit your alerts afterwards, open the Monitoring Dashboard.

Review the activity timeline

Open the Activity Timeline tab, and choose a month in Select Month:. The tab lists the actions taken on the cluster that month, with the time, who did it, and whether it worked. Use it to find out what changed on the cluster, and who changed it.

(Optional) Monitor inside the cluster

The tabs above watch the nodes. To watch pods, containers, and your applications’ own metrics, run a monitoring stack in the cluster. The quickest way is the Monitoring Stack on the cluster’s Marketplace tab. It installs Prometheus, Grafana, and Alertmanager in the namespace kube-prometheus-stack, and you open Grafana through kubectl port-forward; see Use the Monitoring Stack. To install one yourself, the kube-prometheus-stack Helm chart installs Prometheus and Grafana.

Your cluster’s nodes now have graphs, alerts, and a record of every action. If a node reads NotReady, see Troubleshooting Kubernetes.