Swap, RAID, Free space, Node alarms

[DavidePrincipi]

Implement new alarms for clusters with an active subscription.

Proposed solution

The current monitoring script node-monitor is very simple and difficult to improve further. It is time to look at standard solutions, like Netdata or Prometheus (or both) to implement the monitoring of cluster nodes and applications.

Alternative solutions

Implement missing alarms and notifications in the existing script.

See also

• Sub-issues below

[gsanchietti]

The node-monitor has been replaced with a new metrics module that introduces a stack composed by:

• Prometheus to gather the metrics and generate the alerts
• Alertmanager to route the alerts
• alert-proxy to forward alerts to my.nethesis.it and my.nethserver.com

Test case 1: new installation

• Install a new cluster with 2 nodes using the latest testing image
• Verify metrics module is installed and running only on the leader node: look for /home/metrics1 and check that Prometheus, Alertmanager and alert-proxy are running:
  • ss -lanp | grep 9091 for Prometheus
  • ss -lanp | grep 9094 for Alertmanager
  • ss -lanp | grep 9095 for alert-proxy
• Verify node_exporter is running on all nodes: systemctl status node-exporter

Test case 2: update

• Install a new cluster with 2 nodes using the latest stable image
• Install node_exporter module on both nodes
• Update the core on both nodes
• Verify metrics module is installed and running only on the leader node: look for /home/metrics1 and check that Prometheus, Alertmanager and alert-proxy are running:
  • ss -lanp | grep 9091 for Prometheus
  • ss -lanp | grep 9094 for Alertmanager
  • ss -lanp | grep 9095 for alert-proxy
• Verify old node_exporter instances have been removed systemctl status node_exporter*.service
• Verify node_exporter is running on all nodes: systemctl status node-exporter
• Verify node-monitor is not running systemctl status node-monitor

Test case 3: switch leader

• After test case 1 or 2
• Promote the worker node to leader
• Verify metrics module is installed and running only on the leader node: look for /home/metrics1 and check that Prometheus, Alertmanager and alert-proxy are running:
  • ss -lanp | grep 9091 for Prometheus
  • ss -lanp | grep 9094 for Alertmanager
  • ss -lanp | grep 9095 for alert-proxy
• Verify the above services are not running on the worker node: /home/metricsX directory must be gone on the old leader

Test case 4: alert to my.nethesis.it

• After test case 1 or 2
• Disable node_exporter on the worker node: systemctl stop node_exporter
• Verify no alert is sent, inside the logs you should see something like:

  Feb 25 08:48:53 rl1.leader.cluster0.gs.nethserver.net alert-proxy[137405]: ALERT CRITICAL node:offline:1
  Feb 25 08:48:53 rl1.leader.cluster0.gs.nethserver.net alert-proxy[137405]: No auth token, alert not sent
  

• Restart node_exporter on the worker_node: systemctl start node_exporter
• Register the cluster with a valid subscription token
• Disable node_exporter on the worker node: systemctl stop node_exporter
• Verify the alert is received by my.nethesis.it

Test case 4: access to Prometheus

• Enable access to Prometheus: api-cli run module/metrics1/configure-module --data '{"prometheus_path": "prometheus", "grafana_path": ""}'
• Verify Prometheus is accessible from the browser at: https://<node_fqdn>/prometheus

Test case 4: access to Grafana

• Enable access to Grafana: api-cli run module/metrics1/configure-module --data '{"prometheus_path": "prometheus", "grafana_path": "grafana"}'
• Verify Grafana is accessible from the browser at: https://<node_fqdn>/grafana
• Login with default credentials: admin / admin

Test case 5: Grafana dashboard

• After test case 4
• Verify that following dashboards are present and contains data:
  • Loki: it must show metrics from Loki (some charts could be empty)
  • Nodes: hardware metrics of all nodes, nodes can be selected from the top bar
• If an alert is raised, it must be visible under the Alerting section

Test case 6: mail alerts

• After test case 5
• Access the Settings page of the cluster and configure a remote or local mail server to send the notifications
• Configure the mail notification: api-cli run module/metrics1/configure-module --data '{"prometheus_path": "prometheus", "grafana_path": "grafana", "mail_to": ["you@nethesis.it"], "mail_from": "no-reply@nethesis.it", "mail_template": ""}
• Generate an alert
• Verify a mail is sent to the destination (the link to alertmanager inside the mail will point to a non-existing URL)

Test case 7: custom alert

• After test case 4
• Follow instructions on how to create a custom alert: https://github.com/NethServer/ns8-metrics?tab=readme-ov-file#customimze-alert-rules
• Verify the alert is generated

Test case 8: mail notification

• After test case 7
• Follow instructions on how to create a custom mail template: https://github.com/NethServer/ns8-metrics?tab=readme-ov-file#customize-alert-mail-template
• Verify the sent mail uses the new template