Important: This documentation is about an older version. It's relevant only to the release noted, many of the features and functions have been updated or replaced. Please view the current version.

Operator and user guide

Monitor Mimir

About dashboards and alerts requirements

Open source

About Grafana Mimir dashboards and alerts requirements

Grafana Mimir dashboards and alerts require certain labels to exist on metrics scraped from Grafana Mimir. Your Prometheus or Grafana Agent must be configured to add these labels in order for the dashboards and alerts to function. The following table shows the required label names and whether they can be customized when compiling dashboards or alerts from sources.

Label name	Configurable?	Description
`cluster`	Yes	The Kubernetes cluster or datacenter where the Mimir cluster is running. You can configure the cluster label via the `per_cluster_label` field in the mixin configuration.
`namespace`	No	The Kubernetes namespace where the Mimir cluster is running.
`job`	Partially	The Kubernetes namespace and Mimir component in the format `<namespace>/<component>`. When running in monolithic mode, the `<component>` should be `mimir`. When running in microservices mode, the `<component>` should be the name of the specific Mimir component (singular), like `distributor`, `ingester` or `store-gateway`. You cannot configure the label name. But you can configure the regular expressions used to match components with the `job_names` field in the mixin configuration.
`pod`	Yes	The unique identifier of a Mimir replica (eg. Pod ID when running on Kubernetes). You can configure the instance label via the `per_instance_label` field in the mixin configuration.
`instance`	Yes	The unique identifier of the node or machine where the Mimir replica is running (eg. the node when running on Kubernetes). You can configure the node label via the `per_node_label` field in the mixin configuration.

For rules and alerts to function properly, you must configure your Prometheus or Grafana Agent to scrape metrics from Grafana Mimir at an interval of 15s or shorter.

Deployment type

By default, Grafana Mimir dashboards assume Mimir is deployed in containers orchestrated by Kubernetes. If you’re running Mimir on baremental, you should set the configuration field deployment_type: 'baremetal' and re-compile the dashboards.

Job selection

A metric could be exposed by multiple Grafana Mimir components, or even different applications running in the same namespace. To provide accurate dashboards and alerts, we use the job label to select a metric from specific components. A job is a combination of namespace and component, for example <namespace>/ingester.

Pre-compiled dashboards and alerts are shipped with a default configuration. If you compile dashboards and alerts from source, you have the option to customize the regular expression used to select each Mimir component through the job_names field in the mixin config.

Default `job` selection in monolithic mode

When running Grafana Mimir in monolithic mode and using the pre-compiled dashboards and alerts, the job label should be set to <namespace>/mimir.

Default `job` selection in microservices mode

When running Grafana Mimir in microservices mode and using the pre-compiled dashboards and alerts, the job label should be set according to the following table.

Mimir service	Expected `job` label
Distributor	`<namespace>/distributor`
Ingester	`<namespace>/ingester`
Querier	`<namespace>/querier`
Ruler	`<namespace>/ruler`
Query-frontend	`<namespace>/query-frontend`
Query-scheduler	`<namespace>/query-scheduler`
Store-gateway	`<namespace>/store-gateway`
Compactor	`<namespace>/compactor`

Additional resources metrics

The Grafana Mimir dashboards displaying CPU, memory, disk, and network resources utilization require Prometheus metrics scraped from the following endpoints:

For more information about the kubelet metrics and cAdvisor metrics exported by the kubelet, refer to Metrics For Kubernetes System Components.

Metrics from kubelet, kube-state-metrics, and cAdvisor must all have a cluster label with the same value as in the Mimir metrics.

Metrics from node_exporter must all have an instance label on them that has the same value as the instance label on Mimir metrics.

Log labels

The Slow queries dashboard uses a Loki data source with the logs from Grafana Mimir to visualize slow queries. The query-frontend component logs slow queries based on how you configured the -query-frontend.log-queries-longer-than parameter. These logs need to have specific labels in order for the dashboard to work.

Label name	Configurable?	Description
`cluster`	Yes	The Kubernetes cluster or datacenter where the Mimir cluster is running. You can configure the cluster label via the `per_cluster_label` field in the mixin configuration.
`namespace`	No	The Kubernetes namespace where the Mimir cluster is running.
`name`	No	Name of the component. For example, `query-frontend`.