Get started with managed collection

This document describes how to set up Google Cloud Managed Service for Prometheus with managed collection. The setup is a minimal example of working ingestion, using a Prometheus deployment that monitors an example application and stores collected metrics in Monarch.

This document shows you how to do the following:

• Set up your environment and command-line tools.
• Set up managed collection for your cluster.
• Configure a resource for target scraping and metric ingestion.
• Migrate existing prometheus-operator custom resources.

We recommend that you use managed collection; it reduces the complexity of deploying, scaling, sharding, configuring, and maintaining the collectors. Managed collection is supported for GKE and all other Kubernetes environments.

Managed collection runs Prometheus-based collectors as a Daemonset and ensures scalability by only scraping targets on colocated nodes. You configure the collectors with lightweight custom resources to scrape exporters using pull collection, then the collectors push the scraped data to the central datastore Monarch. Google Cloud never directly accesses your cluster to pull or scrape metric data; your collectors push data to Google Cloud. For more information about managed and self-deployed data collection, see Data collection with Managed Service for Prometheus and Ingestion and querying with managed and self-deployed collection.

Before you begin

This section describes the configuration needed for the tasks described in this document.

Set up projects and tools

To use Google Cloud Managed Service for Prometheus, you need the following resources:

• A Google Cloud project with the Cloud Monitoring API enabled.

  • If you don't have a Google Cloud project, then do the following:

    1. In the Google Cloud console, go to New Project:

       Create a New Project

    2. In the Project Name field, enter a name for your project and then click Create.

    3. Go to Billing:

       Go to Billing

    4. Select the project you just created if it isn't already selected at the top of the page.

    5. You are prompted to choose an existing payments profile or to create a new one.

    The Monitoring API is enabled by default for new projects.

  • If you already have a Google Cloud project, then ensure that the Monitoring API is enabled:

    1. Go to APIs & services:

       Go to APIs & services

    2. Select your project.

    3. Click Enable APIs and Services.

    4. Search for "Monitoring".

    5. In the search results, click through to "Cloud Monitoring API".

    6. If "API enabled" is not displayed, then click the Enable button.

• A Kubernetes cluster. If you do not have a Kubernetes cluster, then follow the instructions in the Quickstart for GKE.

You also need the following command-line tools:

• gcloud
• kubectl

The gcloud and kubectl tools are part of the Google Cloud CLI. For information about installing them, see Managing Google Cloud CLI components. To see the gcloud CLI components you have installed, run the following command:

gcloud components list

Configure your environment

To avoid repeatedly entering your project ID or cluster name, perform the following configuration:

• Configure the command-line tools as follows:

  • Configure the gcloud CLI to refer to the ID of your Google Cloud project:

    gcloud config set project PROJECT_ID
    

  • Configure the kubectl CLI to use your cluster:

    kubectl config set-cluster CLUSTER_NAME
    

  For more information about these tools, see the following:

  • gcloud CLI overview
  • kubectl commands

Set up a namespace

Create the NAMESPACE_NAME Kubernetes namespace for resources you create as part of the example application:

kubectl create ns NAMESPACE_NAME

Set up managed collection

You can use managed collection on both GKE and non-GKE Kubernetes clusters.

After managed collection is enabled, the in-cluster components will be running but no metrics are generated yet. PodMonitoring or ClusterPodMonitoring resources are needed by these components to correctly scrape the metrics endpoints. You must either deploy these resources with valid metrics endpoints or enable one of the managed metrics packages, for example, Kube state metrics, built into GKE. For troubleshooting information, see Ingestion-side problems.

Enabling managed collection installs the following components in your cluster:

• The gmp-operator Deployment, which deploys the Kubernetes operator for Managed Service for Prometheus.
• The rule-evaluator Deployment, which is used to configure and run alerting and recording rules.
• The collector DaemonSet, which horizontally scales collection by scraping metrics only from pods running on the same node as each collector.
• The alertmanager StatefulSet, which is configured to send triggered alerts to your preferred notification channels.

For reference documentation about the Managed Service for Prometheus operator, see the manifests page.

Enable managed collection: GKE

Managed collection is enabled by default for the following:

• GKE Autopilot clusters running GKE version 1.25 or greater.

• GKE Standard clusters running GKE version 1.27 or greater. You can override this default when creating the cluster; see Disable managed collection.

If you are running in a GKE environment that does not enable managed collection by default, then see Enable managed collection manually.

Managed collection on GKE is automatically upgraded when new in-cluster component versions are released.

Managed collection on GKE uses permissions granted to the default Compute Engine service account. If you have a policy that modifies the standard permissions on the default node service account, you might need to add the Monitoring Metric Writer role to continue.

Enable managed collection manually

If you are running in a GKE environment that does not enable managed collection by default, then you can enable managed collection by using the following:

• The GKE Clusters dashboard in Cloud Monitoring.
• The Kubernetes Engine page in the Google Cloud console.
• The Google Cloud CLI. To use the gcloud CLI, you must be running GKE version 1.21.4-gke.300 or newer.

• Terraform for Google Kubernetes Engine. To use Terraform to enable Managed Service for Prometheus, you must be running GKE version 1.21.4-gke.300 or newer.

gcloud config set project PROJECT_ID

gcloud container clusters create CLUSTER_NAME --zone ZONE --enable-managed-prometheus

Disable managed collection

If you want to disable managed collection on your clusters, then you can use one of the following methods:

gcloud config set project PROJECT_ID

gcloud container clusters create CLUSTER_NAME --zone ZONE --no-enable-managed-prometheus

Enable managed collection: non-GKE Kubernetes

If you are running in a non-GKE environment, then you can enable managed collection using the following:

• The kubectl CLI.

• The bundled solution included in GKE Enterprise deployments running version 1.12 or newer.

kubectl apply -f https://raw.githubusercontent.com/GoogleCloudPlatform/prometheus-engine/v0.13.0/manifests/setup.yaml

kubectl apply -f https://raw.githubusercontent.com/GoogleCloudPlatform/prometheus-engine/v0.13.0/manifests/operator.yaml

Deploy the example application

The example application emits the example_requests_total counter metric and the example_random_numbers histogram metric (among others) on its metrics port. The manifest for the application defines three replicas.

To deploy the example application, run the following command:

kubectl -n NAMESPACE_NAME apply -f https://raw.githubusercontent.com/GoogleCloudPlatform/prometheus-engine/v0.13.0/examples/example-app.yaml

Configure a PodMonitoring resource

To ingest the metric data emitted by the example application, Managed Service for Prometheus uses target scraping. Target scraping and metrics ingestion are configured using Kubernetes custom resources. The managed service uses PodMonitoring custom resources (CRs).

A PodMonitoring CR scrapes targets only in the namespace the CR is deployed in. To scrape targets in multiple namespaces, deploy the same PodMonitoring CR in each namespace. You can verify the PodMonitoring resource is installed in the intended namespace by running kubectl get podmonitoring -A.

For reference documentation about all the Managed Service for Prometheus CRs, see the prometheus-engine/doc/api reference.

The following manifest defines a PodMonitoring resource, prom-example, in the NAMESPACE_NAME namespace. The resource uses a Kubernetes label selector to find all pods in the namespace that have the label app.kubernetes.io/name with the value prom-example. The matching pods are scraped on a port named metrics, every 30 seconds, on the /metrics HTTP path.

apiVersion: monitoring.googleapis.com/v1
kind: PodMonitoring
metadata:
  name: prom-example
spec:
  selector:
    matchLabels:
      app.kubernetes.io/name: prom-example
  endpoints:
  - port: metrics
    interval: 30s

To apply this resource, run the following command:

kubectl -n NAMESPACE_NAME apply -f https://raw.githubusercontent.com/GoogleCloudPlatform/prometheus-engine/v0.13.0/examples/pod-monitoring.yaml

Your managed collector is now scraping the matching pods. You can view the status of your scrape target by enabling the target status feature.

To configure horizontal collection that applies to a range of pods across all namespaces, use the ClusterPodMonitoring resource. The ClusterPodMonitoring resource provides the same interface as the PodMonitoring resource but does not limit discovered pods to a given namespace.

If you are running on GKE, then you can do the following:

• To query the metrics ingested by the example application using PromQL in Cloud Monitoring, see Query using Cloud Monitoring.
• To query the metrics ingested by the example application using Grafana, see Query using Grafana or any Prometheus API consumer.
• To learn about filtering exported metrics and adapting your prom-operator resources, see Additional topics for managed collection.

If you are running outside of GKE, then you need to create a service account and authorize it to write your metric data, as described in the following section.

Provide credentials explicitly

When running on GKE, the collecting Prometheus server automatically retrieves credentials from the environment based on the node's service account. In non-GKE Kubernetes clusters, credentials must be explicitly provided through the OperatorConfig resource in the gmp-public namespace.

1. Set the context to your target project:

   gcloud config set project PROJECT_ID
   

2. Create a service account:

   gcloud iam service-accounts create gmp-test-sa
   

3. Grant the required permissions to the service account:

   gcloud projects add-iam-policy-binding PROJECT_ID\
     --member=serviceAccount:gmp-test-sa@PROJECT_ID.iam.gserviceaccount.com \
     --role=roles/monitoring.metricWriter
   

4. Create and download a key for the service account:

   gcloud iam service-accounts keys create gmp-test-sa-key.json \
     --iam-account=gmp-test-sa@PROJECT_ID.iam.gserviceaccount.com
   

5. Add the key file as a secret to your non-GKE cluster:

   kubectl -n gmp-public create secret generic gmp-test-sa \
     --from-file=key.json=gmp-test-sa-key.json
   

6. Open the OperatorConfig resource for editing:

   kubectl -n gmp-public edit operatorconfig config
   

   1. Add the text shown in bold to the resource:

      apiVersion: monitoring.googleapis.com/v1
      kind: OperatorConfig
      metadata:
        namespace: gmp-public
        name: config
      collection:
        credentials:
          name: gmp-test-sa
          key: key.json
      

      Make sure you also add these credentials to the rules section so that managed rule evaluation works.

   2. Save the file and close the editor. After the change is applied, the pods are re-created and start authenticating to the metric backend with the given service account.

   Additional topics for managed collection

   This section describes how to do the following:

   • Enable the target status feature for easier debugging.
   • Configure target scraping using Terraform.
   • Filter the data you export to the managed service.
   • Scrape Kubelet and cAdvisor metrics.
   • Convert your existing prom-operator resources for use with the managed service.
   • Run managed collection outside of GKE.

   Enabling the target status feature

   Managed Service for Prometheus provides a way to check whether your targets are being properly discovered and scraped by the collectors. This target status report is meant to be a tool for debugging acute problems. We strongly recommend only enabling this feature to investigate immediate issues. Leaving target status reporting on in large clusters might cause the operator to run out of memory and crash loop.

   You can check the status of your targets in your PodMonitoring or ClusterPodMonitoring resources by setting the features.targetStatus.enabled value within the OperatorConfig resource to true, as shown in the following:

       apiVersion: monitoring.googleapis.com/v1
       kind: OperatorConfig
       metadata:
         namespace: gmp-public
         name: config
       features:
         targetStatus:
           enabled: true
   

   After a few seconds, the Status.Endpoint Statuses field appears on every valid PodMonitoring or ClusterPodMonitoring resource, when configured.

   If you have a PodMonitoring resource with the name prom-example in the NAMESPACE_NAME namespace, then you can check the status by running the following command:

   kubectl -n NAMESPACE_NAME describe podmonitorings/prom-example
   

   The output looks like the following:

   API Version:  monitoring.googleapis.com/v1
   Kind:         PodMonitoring
   ...
   Status:
     Conditions:
       ...
       Status:                True
       Type:                  ConfigurationCreateSuccess
     Endpoint Statuses:
       Active Targets:       3
       Collectors Fraction:  1
       Last Update Time:     2023-08-02T12:24:26Z
       Name:                 PodMonitoring/custom/prom-example/metrics
       Sample Groups:
         Count:  3
         Sample Targets:
           Health:  up
           Labels:
             Cluster:                     CLUSTER_NAME
             Container:                   prom-example
             Instance:                    prom-example-589ddf7f7f-hcnpt:metrics
             Job:                         prom-example
             Location:                    REGION
             Namespace:                   NAMESPACE_NAME
             Pod:                         prom-example-589ddf7f7f-hcnpt
             project_id:                  PROJECT_ID
           Last Scrape Duration Seconds:  0.020206416
           Health:                        up
           Labels:
             ...
           Last Scrape Duration Seconds:  0.054189485
           Health:                        up
           Labels:
             ...
           Last Scrape Duration Seconds:  0.006224887
   

   The output includes the following status fields:

   • Status.Conditions.Status is true when Managed Service for Prometheus acknowledges and processes the PodMonitoring or ClusterPodMonitoring.
   • Status.Endpoint Statuses.Active Targets shows the number of scrape targets that Managed Service for Prometheus counts on all collectors for this PodMonitoring resource. In the example application, the prom-example deployment has three replicas with a single metric target, so the value is 3. If there are unhealthy targets, the Status.Endpoint Statuses.Unhealthy Targets field appears.
   • Status.Endpoint Statuses.Collectors Fraction shows a value of 1 (meaning 100%) if all of the managed collectors are reachable by Managed Service for Prometheus.
   • Status.Endpoint Statuses.Last Update Time shows the last updated time. When the last update time is significantly longer than your desired scrape interval time, the difference might indicate issues with your target or cluster.
   • Status.Endpoint Statuses.Sample Groups field shows sample targets grouped by common target labels injected by the collector. This value is useful for debugging situations where your targets are not discovered. If all targets are healthy and being collected, then the expected value for the Health field is up and the value for the Last Scrape Duration Seconds field is the usual duration for a typical target.

   For more information about these fields, see the Managed Service for Prometheus API document.

   Any of the following might indicate a problem wth your configuration:

   • There is no Status.Endpoint Statuses field in your PodMonitoring resource.
   • The value of the Last Scrape Duration Seconds field is too old.
   • You see too few targets.
   • The value of the Health field indicates that the target is down.

   For more information about debugging target discovery issues, see Ingestion-side problems in the troubleshooting documentation.

   Configuring an authorized scrape endpoint

   If your scrape target requires authorization, you can set up the collector to use the correct authorization type and provide any relevant secrets.

   Google Cloud Managed Service for Prometheus supports the following authorization types:

   • mTLS
   • BasicAuth
   • HTTP Authorization Header
   • OAuth 2

   mTLS

   mTLS is commonly configured within zero trust environments, such as Istio service mesh or Cloud Service Mesh.

   To enable scraping endpoints secured using mTLS, set the Spec.Endpoints[].Scheme field in your PodMonitoring resource to https. While not recommended, you can set the Spec.Endpoints[].insecureSkipVerify field in your PodMonitoring resource to true to skip verifying the certificate authority. Alternatively, you can configure Managed Service for Prometheus to load certificates and keys from secret resources.

   For example, the following Secret resource contains keys for the client (cert), private key (key), and certificate authority (ca) certificates:

   kind: Secret
   metadata:
     name: secret-example
   stringData:
     cert: ********
     key: ********
     ca: ********
   

   Grant the Managed Service for Prometheus collector permission to access that Secret resource:

   apiVersion: rbac.authorization.k8s.io/v1
   kind: Role
   metadata:
     name: secret-example-read
   rules:
   - resources:
     - secrets
     apiGroups: [""]
     verbs: ["get", "list", "watch"]
     resourceNames: ["secret-example"]
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: RoleBinding
   metadata:
     name: gmp-system:collector:secret-example-read
     namespace: default
   roleRef:
     name: secret-example-read
     kind: Role
     apiGroup: rbac.authorization.k8s.io
   subjects:
   - name: collector
     namespace: gmp-system
     kind: ServiceAccount
   

   On GKE Autopilot clusters, this looks like:

   apiVersion: rbac.authorization.k8s.io/v1
   kind: Role
   metadata:
     name: secret-example-read
   rules:
   - resources:
     - secrets
     apiGroups: [""]
     verbs: ["get", "list", "watch"]
     resourceNames: ["secret-example"]
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: RoleBinding
   metadata:
     name: gmp-system:collector:secret-example-read
     namespace: default
   roleRef:
     name: secret-example-read
     kind: Role
     apiGroup: rbac.authorization.k8s.io
   subjects:
   - name: collector
     namespace: gke-gmp-system
     kind: ServiceAccount
   

   To configure a PodMonitoring resource that uses the prior Secret resource, modify your resource to add a scheme and tls section:

   apiVersion: monitoring.googleapis.com/v1
   kind: PodMonitoring
   metadata:
     name: prom-example
   spec:
     selector:
       matchLabels:
         app.kubernetes.io/name: prom-example
     endpoints:
     - port: metrics
       interval: 30s
       scheme: https
       tls:
         ca:
           secret:
             name: secret-example
             key: ca
         cert:
           secret:
             name: secret-example
             key: cert
         key:
           secret:
             name: secret-example
             key: key
   

   For reference documentation about all the Managed Service for Prometheus mTLS options, see the API reference documentation.

   BasicAuth

   To enable scraping endpoints secured using BasicAuth, set the Spec.Endpoints[].BasicAuth field in your PodMonitoring resource with your username and password. For other HTTP Authorization Header types, see HTTP Authorization Header.

   For example, the following Secret resource contains a key to store the password:

   kind: Secret
   metadata:
     name: secret-example
   stringData:
     password: ********
   

   Grant the Managed Service for Prometheus collector permission to access that Secret resource:

   apiVersion: rbac.authorization.k8s.io/v1
   kind: Role
   metadata:
     name: secret-example-read
   rules:
   - resources:
     - secrets
     apiGroups: [""]
     verbs: ["get", "list", "watch"]
     resourceNames: ["secret-example"]
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: RoleBinding
   metadata:
     name: gmp-system:collector:secret-example-read
     namespace: default
   roleRef:
     name: secret-example-read
     kind: Role
     apiGroup: rbac.authorization.k8s.io
   subjects:
   - name: collector
     namespace: gmp-system
     kind: ServiceAccount
   

   On GKE Autopilot clusters, this looks like:

   apiVersion: rbac.authorization.k8s.io/v1
   kind: Role
   metadata:
     name: secret-example-read
   rules:
   - resources:
     - secrets
     apiGroups: [""]
     verbs: ["get", "list", "watch"]
     resourceNames: ["secret-example"]
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: RoleBinding
   metadata:
     name: gmp-system:collector:secret-example-read
     namespace: default
   roleRef:
     name: secret-example-read
     kind: Role
     apiGroup: rbac.authorization.k8s.io
   subjects:
   - name: collector
     namespace: gke-gmp-system
     kind: ServiceAccount
   

   To configure a PodMonitoring resource that uses the prior Secret resource and a username of foo, modify your resource to add a basicAuth section:

   apiVersion: monitoring.googleapis.com/v1
   kind: PodMonitoring
   metadata:
     name: prom-example
   spec:
     selector:
       matchLabels:
         app.kubernetes.io/name: prom-example
     endpoints:
     - port: metrics
       interval: 30s
       basicAuth:
         username: foo
         password:
           secret:
             name: secret-example
             key: password
   

   For reference documentation about all the Managed Service for Prometheus BasicAuth options, see the API reference documentation.

   HTTP Authorization Header

   To enable scraping endpoints secured using HTTP Authorization Headers, set the Spec.Endpoints[].Authorization field in your PodMonitoring resource with the type and credentials. For BasicAuth endpoints, use the BasicAuth configuration instead.

   For example, the following Secret resource contains a key to store the credentials:

   kind: Secret
   metadata:
     name: secret-example
   stringData:
     credentials: ********
   

   Grant the Managed Service for Prometheus collector permission to access that Secret resource:

   apiVersion: rbac.authorization.k8s.io/v1
   kind: Role
   metadata:
     name: secret-example-read
   rules:
   - resources:
     - secrets
     apiGroups: [""]
     verbs: ["get", "list", "watch"]
     resourceNames: ["secret-example"]
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: RoleBinding
   metadata:
     name: gmp-system:collector:secret-example-read
     namespace: default
   roleRef:
     name: secret-example-read
     kind: Role
     apiGroup: rbac.authorization.k8s.io
   subjects:
   - name: collector
     namespace: gmp-system
     kind: ServiceAccount
   

   On GKE Autopilot clusters, this looks like:

   apiVersion: rbac.authorization.k8s.io/v1
   kind: Role
   metadata:
     name: secret-example-read
   rules:
   - resources:
     - secrets
     apiGroups: [""]
     verbs: ["get", "list", "watch"]
     resourceNames: ["secret-example"]
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: RoleBinding
   metadata:
     name: gmp-system:collector:secret-example-read
     namespace: default
   roleRef:
     name: secret-example-read
     kind: Role
     apiGroup: rbac.authorization.k8s.io
   subjects:
   - name: collector
     namespace: gke-gmp-system
     kind: ServiceAccount
   

   To configure a PodMonitoring resource that uses the prior Secret resource and a type of Bearer, modify your resource to add an authorization section:

   apiVersion: monitoring.googleapis.com/v1
   kind: PodMonitoring
   metadata:
     name: prom-example
   spec:
     selector:
       matchLabels:
         app.kubernetes.io/name: prom-example
     endpoints:
     - port: metrics
       interval: 30s
       authorization:
         type: Bearer
         credentials:
           secret:
             name: secret-example
             key: credentials
   

   For reference documentation about all the Managed Service for Prometheus HTTP Authorization Header options, see the API reference documentation.

   OAuth 2

   To enable scraping endpoints secured using OAuth 2, you must set the Spec.Endpoints[].OAuth2 field in your PodMonitoring resource.

   For example, the following Secret resource contains a key to store the client secret:

   kind: Secret
   metadata:
     name: secret-example
   stringData:
     clientSecret: ********
   

   Grant the Managed Service for Prometheus collector permission to access that Secret resource:

   apiVersion: rbac.authorization.k8s.io/v1
   kind: Role
   metadata:
     name: secret-example-read
   rules:
   - resources:
     - secrets
     apiGroups: [""]
     verbs: ["get", "list", "watch"]
     resourceNames: ["secret-example"]
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: RoleBinding
   metadata:
     name: gmp-system:collector:secret-example-read
     namespace: default
   roleRef:
     name: secret-example-read
     kind: Role
     apiGroup: rbac.authorization.k8s.io
   subjects:
   - name: collector
     namespace: gmp-system
     kind: ServiceAccount
   

   On GKE Autopilot clusters, this looks like:

   apiVersion: rbac.authorization.k8s.io/v1
   kind: Role
   metadata:
     name: secret-example-read
   rules:
   - resources:
     - secrets
     apiGroups: [""]
     verbs: ["get", "list", "watch"]
     resourceNames: ["secret-example"]
   ---
   apiVersion: rbac.authorization.k8s.io/v1
   kind: RoleBinding
   metadata:
     name: gmp-system:collector:secret-example-read
     namespace: default
   roleRef:
     name: secret-example-read
     kind: Role
     apiGroup: rbac.authorization.k8s.io
   subjects:
   - name: collector
     namespace: gke-gmp-system
     kind: ServiceAccount
   

   To configure a PodMonitoring resource that uses the prior Secret resource with a client ID of foo and token URL of example.com/token, modify your resource to add an oauth2 section:

   apiVersion: monitoring.googleapis.com/v1
   kind: PodMonitoring
   metadata:
     name: prom-example
   spec:
     selector:
       matchLabels:
         app.kubernetes.io/name: prom-example
     endpoints:
     - port: metrics
       interval: 30s
       oauth2:
         clientID: foo
         clientSecret:
           secret:
             name: secret-example
             key: password
         tokenURL: example.com/token
   

   For reference documentation about all the Managed Service for Prometheus OAuth 2 options, see the API reference documentation.

   Configuring target scraping using Terraform

   You can automate the creation and management of PodMonitoring and ClusterPodMonitoring resources by using the kubernetes_manifest Terraform resource type or the kubectl_manifest Terraform resource type, either of which lets you specify arbitrary custom resources.

   For general information about using Google Cloud with Terraform, see Terraform with Google Cloud.

   Filter exported metrics

   If you collect a lot of data, you might want to prevent some time series from being sent to Managed Service for Prometheus to keep down costs. You can do this by using Prometheus relabeling rules with a keep action for an allowlist or a drop action for a denylist. For managed collection, this rule goes in the metricRelabeling section of your PodMonitoring or ClusterPodMonitoring resource.

   For example, the following metric relabeling rule will filter out any metric that begins with foo_bar_, foo_baz_, or foo_qux_:

     metricRelabeling:
     - action: drop
       regex: foo_(bar|baz|qux)_.+
       sourceLabels: [__name__]
   

   The Cloud Monitoring Metrics Management page provides information that can help you control the amount you spend on billable metrics without affecting observability. The Metrics Management page reports the following information:

   • Ingestion volumes for both byte- and sample-based billing, across metric domains and for individual metrics.
   • Data about labels and cardinality of metrics.
   • Number of reads for each metric.
   • Use of metrics in alerting policies and custom dashboards.
   • Rate of metric-write errors.

   You can also use the Metrics Management to exclude unneeded metrics, eliminating the cost of ingesting them. For more information about the Metrics Management page, see View and manage metric usage.

   For additional suggestions on how to lower your costs, see Cost controls and attribution.

   Scraping Kubelet and cAdvisor metrics

   The Kubelet exposes metrics about itself as well as cAdvisor metrics about containers running on its node. You can configure managed collection to scrape Kubelet and cAdvisor metrics by editing the OperatorConfig resource. For instructions, see the exporter documentation for Kubelet and cAdvisor.

   Convert existing prometheus-operator resources

   You can usually convert your existing prometheus-operator resources to Managed Service for Prometheus managed collection PodMonitoring and ClusterPodMonitoring resources.

   For example, the ServiceMonitor resource defines monitoring for a set of services. The PodMonitoring resource serves a subset of the fields served by the ServiceMonitor resource. You can convert a ServiceMonitor CR to a PodMonitoring CR by mapping the fields as described in the following table:

   monitoring.coreos.com/v1 ServiceMonitor	Compatibility	monitoring.googleapis.com/v1 PodMonitoring
   .ServiceMonitorSpec.Selector	Identical	.PodMonitoringSpec.Selector
   .ServiceMonitorSpec.Endpoints[]	.TargetPort maps to .Port .Path: compatible .Interval: compatible .Timeout: compatible	.PodMonitoringSpec.Endpoints[]
   .ServiceMonitorSpec.TargetLabels	PodMonitor must specify: .FromPod[].From pod label .FromPod[].To target label	.PodMonitoringSpec.TargetLabels

   The following is a sample ServiceMonitor CR; the content in bold type is replaced in the conversion, and the content in italic type maps directly:

   apiVersion: monitoring.coreos.com/v1
   kind: ServiceMonitor
   metadata:
     name: example-app
   spec:
     selector:
       matchLabels:
         app: example-app
     endpoints:
     - targetPort: web
       path: /stats
       interval: 30s
     targetLabels:
     - foo
   

   The following is the analogous PodMonitoring CR, assuming that your service and its pods are labeled with app=example-app. If this assumption does not apply, then you need to use the label selectors of the underlying Service resource.

   The content in bold type has been replaced in the conversion:

   apiVersion: monitoring.googleapis.com/v1
   kind: PodMonitoring
   metadata:
     name: example-app
   spec:
     selector:
       matchLabels:
         app: example-app
     endpoints:
     - port: web
       path: /stats
       interval: 30s
     targetLabels:
       fromPod:
       - from: foo # pod label from example-app Service pods.
         to: foo
   

   You can always continue to use your existing prometheus-operator resources and deployment configs by using self-deployed collectors instead of managed collectors. You can query metrics sent from both collector types, so you might want to use self-deployed collectors for your existing Prometheus deployments while using managed collectors for new Prometheus deployments.

   Reserved labels

   Managed Service for Prometheus automatically adds the following labels to all metrics collected. These labels are used to uniquely identify a resource in Monarch:

   • project_id: The identifier of the Google Cloud project associated with your metric.
   • location: The physical location (Google Cloud region) where the data is stored. This value is typically the region of your GKE cluster. If data is collected from an AWS or on-premises deployment, then the value might be the closest Google Cloud region.
   • cluster: The name of the Kubernetes cluster associated with your metric.
   • namespace: The name of the Kubernetes namespace associated with your metric.
   • job: The job label of the Prometheus target, if known; might be empty for rule-evaluation results.
   • instance: The instance label of the Prometheus target, if known; might be empty for rule-evaluation results.

   While not recommended when running on Google Kubernetes Engine, you can override the project_id, location, and cluster labels by adding them as args to the Deployment resource within operator.yaml. If you use any reserved labels as metric labels, Managed Service for Prometheus automatically relabels them by adding the prefix exported_. This behavior matches how upstream Prometheus handles conflicts with reserved labels.

   Compress configurations

   If you have many PodMonitoring resources, you might run out of ConfigMap space. To fix this, enable gzip compression in your OperatorConfig resource:

     apiVersion: monitoring.googleapis.com/v1
     kind: OperatorConfig
     metadata:
       namespace: gmp-public
       name: config
     features:
       config:
         compression: gzip
   

   Enable vertical pod autoscaling (VPA) for managed collection

   If you are encountering Out of Memory (OOM) errors for the collector pods in your cluster or if the default resource requests and limits for the collectors otherwise don't meet your needs, then you can use vertical pod autoscaling to dynamically allocate resources.

   When you set the field scaling.vpa.enabled: true on the OperatorConfig resource, the operator deploys a VerticalPodAutoscaling manifest in the cluster that allows the resource requests and limits of the collector pods to be set automatically, based on usage.

   To enable VPA for collector pods in Managed Service for Prometheus, run the following command:

   kubectl -n gmp-public patch operatorconfig/config -p '{"scaling":{"vpa":{"enabled":true}}}' --type=merge
   

   If the command completes successfully, then the operator sets up vertical pod autoscaling for the collector pods. Out Of Memory errors result in an immediate increase to the resource limits. If there are no OOM errors, then the first adjustment to the resource requests and limits of the collector pods typically occurs within 24 hours.

   You might receive this error when attempting to enable VPA:

   vertical pod autoscaling is not available - install vpa support and restart the operator

   To resolve this error, you need to first enable vertical pod autoscaling at the cluster level:

   1. Go to the Kubernetes Engine - Clusters page in the Google Cloud console.

      In the Google Cloud console, go to the Kubernetes clusters page:

      Go to Kubernetes clusters

      If you use the search bar to find this page, then select the result whose subheading is Kubernetes Engine.

   2. Select the cluster you want to modify.

   3. In the Automation section, edit the value of the Vertical Pod Autoscaling option.

   4. Select the Enable Vertical Pod Autoscaling checkbox, and then click Save changes. This change restarts your cluster. The operator restarts as a part of this process.

   5. Retry the following command: kubectl -n gmp-public patch operatorconfig/config -p '{"scaling":{"vpa":{"enabled":true}}}' --type=merge to enable VPA for Managed Service for Prometheus.

   To confirm that the OperatorConfig resource is edited successfully, open it using the command kubectl -n gmp-public edit operatorconfig config. If successful, your OperatorConfig includes the following section in bold:

   apiVersion: monitoring.googleapis.com/v1
   kind: OperatorConfig
   metadata:
     namespace: gmp-public
     name: config
   scaling:
     vpa:
       enabled: true
   

   Vertical pod autoscaling works best when ingesting steady numbers of samples, divided equally across nodes. If the metrics load is irregular or spiky, or if metrics load varies greatly between nodes, VPA might not be an efficient solution.

   For more information, see vertical pod autoscaling in GKE.

   Teardown

   To disable managed collection deployed using gcloud or the GKE UI, you can do either of the following:

   • Run the following command:

     gcloud container clusters update CLUSTER_NAME --disable-managed-prometheus
     

   • Use the GKE UI:

     1. Select Kubernetes Engine in the Google Cloud console, then select Clusters.

     2. Locate the cluster for which you want to disable managed collection and click its name.

     3. On the Details tab, scroll down to Features and change the state to Disabled by using the edit button.

   To disable managed collection deployed by using Terraform, specify enabled = false in the managed_prometheus section of the google_container_cluster resource.

   To disable managed collection deployed by using kubectl, run the following command:

   kubectl delete -f https://raw.githubusercontent.com/GoogleCloudPlatform/prometheus-engine/v0.13.0/manifests/operator.yaml
   

   Disabling managed collection causes your cluster to stop sending new data to Managed Service for Prometheus. Taking this action does not delete any existing metrics data already stored in the system.

   Disabling managed collection also deletes the gmp-public namespace and any resources within it, including any exporters installed in that namespace.

   Run managed collection outside of GKE

   In GKE environments, you can run managed collection without further configuration. In other Kubernetes environments, you need to explicitly provide credentials, a project-id value to contain your metrics, a location value (Google Cloud region) where your metrics will be stored, and a cluster value to save the name of the cluster in which the collector is running.

   As gcloud does not work outside of Google Cloud environments, you need to deploy using kubectl instead. Unlike with gcloud, deploying managed collection using kubectl does not automatically upgrade your cluster when a new version is available. Remember to watch the releases page for new versions and manually upgrade by re-running the kubectl commands with the new version.

   You can provide a service account key by modifying the OperatorConfig resource within operator.yaml as described in Provide credentials explicitly. You can provide project-id, location, and cluster values by adding them as args to the Deployment resource within operator.yaml.

   We recommend choosing project-id based on your planned tenancy model for reads. Pick a project to store metrics in based on how you plan to organize reads later with metrics scopes. If you don't care, you can put everything into one project.

   For location, we recommend choosing the nearest Google Cloud region to your deployment. The further the chosen Google Cloud region is from your deployment, the more write latency you'll have and the more you'll be affected by potential networking issues. You might want to consult this list of regions across multiple clouds. If you don't care, you can put everything into one Google Cloud region. You can't use global as your location.

   For cluster, we recommend choosing the name of the cluster in which the operator is deployed.

   When properly configured, your OperatorConfig should look like this:

       apiVersion: monitoring.googleapis.com/v1
       kind: OperatorConfig
       metadata:
         namespace: gmp-public
         name: config
       collection:
         credentials:
           name: gmp-test-sa
           key: key.json
       rules:
         credentials:
           name: gmp-test-sa
           key: key.json
   

   And your Deployment resource should look like this:

   apiVersion: apps/v1
   kind: Deployment
   ...
   spec:
     ...
     template:
       ...
       spec:
         ...
         containers:
         - name: operator
           ...
           args:
           - ...
           - "--project-id=PROJECT_ID"
           - "--cluster=CLUSTER_NAME"
           - "--location=REGION"
   

   This example assumes you have set the REGION variable to a value like us-central1, for example.

   Running Managed Service for Prometheus outside of Google Cloud incurs data transfer fees. There are fees to transfer data into Google Cloud, and you might incur fees to transfer data out of another cloud. You can minimize these costs by enabling gzip compression over the wire through the OperatorConfig. Add the text shown in bold to the resource:

       apiVersion: monitoring.googleapis.com/v1
       kind: OperatorConfig
       metadata:
         namespace: gmp-public
         name: config
       collection:
         compression: gzip
         ...
   

   Further reading on managed collection custom resources

   For reference documentation about all the Managed Service for Prometheus custom resources, see the prometheus-engine/doc/api reference.

   What's next

   • Use PromQL in Cloud Monitoring to query Prometheus metrics.
   • Use Grafana to query Prometheus metrics.
   • Use PromQL alerts in Cloud Monitoring.
   • Set up managed rule evaluation.
   • Set up commonly used exporters.