Usage recommendations#

This section provides guidelines for using the monitoring service.

Prometheus#

This section provides guidelines for using Prometheus monitoring service.

Service launch options#

  • Standalone caching instance in the selected Availability Zone

Using labels#

When configuring and using a Prometheus-based monitoring service, labels are widely used. All metrics generated by exporters already contain a number of labels, which can be either common for a service or node or specific. To see which labels are assigned to metrics, go to the “Graph” section of the Prometheus web interface, or make an HTTP GET request to the endpoint of a specific exporter.

Labels are also added to alerts (for example, severity label indicates the failure severity). In addition, you can add your own labels by specifying them in the settings of PaaS services or in the scrape job. For example, when deploying a MongoDB PaaS service, you can assign it a label with the key env and the value testing to indicate that it belongs to the test environment.

Labels are useful when configuring notification rules and disabling notifications (silence) in the Alertmanager interface, as well as when configuring matchers for a notification channel in the cloud web interface. For example, to send email alerts triggered by a critical failure of the MongoDB service related to the test environment, you need to make the following steps:

  • create a notification channel of the email type, for example, with the name paasalerts;

  • create a route in which you specify paasalerts as the notification channel and enter matchers:

    • severity: critical

    • env: testing

    • service: mongodb

All metrics and warnings generated by the cloud monitoring service contain the following labels:

  • paas_type – PaaS service type in the lower case, for example mysql, pgsql, mongodb, etc.

  • paas_name – the unique name of the deployed PaaS service, which is specified by the user at the creation stage;

  • node – has the format <hostname>@<paas_name>, where <hostname> is the source virtual machine name.

Alarms must also contain the following labels:

  • severity – the warning severity can have the following values: critical, warning, info

  • alertname – the alert name.

You can use this predefined label set when configuring route selection criteria.

Connection to the service#

Once the monitoring service is deployed, three services are available on the corresponding virtual machine: Prometheus, Alertmanager, and Grafana. Each of them has its own web interface running on the corresponding port:

  • Prometheus – tcp/9090;

  • Alertmanager – tcp/9093;

  • Grafana – tcp/3000.

To open the required web interface, type http://<vm_address>:<port> in your browser, where vm_address is the IP address of the instance with the monitoring service installed.

The Prometheus and Alertmanager web interfaces do not require authentication, so it is better not to open their ports to everyone. To work with these services over the Internet, it is recommended to either restrict access to them by using security groups, or install a reverse proxy in front of them for client authentication or address filtering.

The default username and password for logging in to Grafana is admin/admin. Once the service is deployed, you must change your password the first time you log in. Although Grafana requires authentication, we also recommend using whitelist to restrict access to it.