Stefan Reimer e3233074c1 feat: kube-prometheus-stack version upgrade incl. latest state metrics and dashboards for the metrics systems themselves, fix for gathering all prometheus rules across all namespaces

2021-06-01 18:40:34 +02:00

24 KiB

Raw Blame History

kube-prometheus-stack

Installs the kube-prometheus stack, a collection of Kubernetes manifests, Grafana dashboards, and Prometheus rules combined with documentation and scripts to provide easy to operate end-to-end Kubernetes cluster monitoring with Prometheus using the Prometheus Operator.

See the kube-prometheus README for details about components, dashboards, and alerts.

Note: This chart was formerly named prometheus-operator chart, now renamed to more clearly reflect that it installs the kube-prometheus project stack, within which Prometheus Operator is only one component.

Prerequisites

Kubernetes 1.16+
Helm 3+

Get Repo Info

helm repo add prometheus-community https://prometheus-community.github.io/helm-charts
helm repo update

See helm repo for command documentation.

Install Chart

# Helm
$ helm install [RELEASE_NAME] prometheus-community/kube-prometheus-stack

See configuration below.

See helm install for command documentation.

Dependencies

By default this chart installs additional, dependent charts:

To disable dependencies during installation, see multiple releases below.

See helm dependency for command documentation.

Uninstall Chart

# Helm
$ helm uninstall [RELEASE_NAME]

This removes all the Kubernetes components associated with the chart and deletes the release.

See helm uninstall for command documentation.

CRDs created by this chart are not removed by default and should be manually cleaned up:

kubectl delete crd alertmanagerconfigs.monitoring.coreos.com
kubectl delete crd alertmanagers.monitoring.coreos.com
kubectl delete crd podmonitors.monitoring.coreos.com
kubectl delete crd probes.monitoring.coreos.com
kubectl delete crd prometheuses.monitoring.coreos.com
kubectl delete crd prometheusrules.monitoring.coreos.com
kubectl delete crd servicemonitors.monitoring.coreos.com
kubectl delete crd thanosrulers.monitoring.coreos.com

Upgrading Chart

# Helm
$ helm upgrade [RELEASE_NAME] prometheus-community/kube-prometheus-stack

With Helm v3, CRDs created by this chart are not updated by default and should be manually updated. Consult also the Helm Documentation on CRDs.

See helm upgrade for command documentation.

Upgrading an existing Release to a new major version

A major chart version change (like v1.2.3 -> v2.0.0) indicates that there is an incompatible breaking change needing manual actions.

From 15.x to 16.x

Version 16 upgrades kube-state-metrics to v2.0.0. This includes changed command-line arguments and removed metrics, see this blog post. This version also removes Grafana dashboards that supported Kubernetes 1.14 or earlier.

From 14.x to 15.x

Version 15 upgrades prometheus-operator from 0.46.x to 0.47.x. Helm does not automatically upgrade or install new CRDs on a chart upgrade, so you have to install the CRDs manually before updating:

kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.47.0/example/prometheus-operator-crd/monitoring.coreos.com_alertmanagerconfigs.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.47.0/example/prometheus-operator-crd/monitoring.coreos.com_alertmanagers.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.47.0/example/prometheus-operator-crd/monitoring.coreos.com_podmonitors.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.47.0/example/prometheus-operator-crd/monitoring.coreos.com_probes.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.47.0/example/prometheus-operator-crd/monitoring.coreos.com_prometheuses.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.47.0/example/prometheus-operator-crd/monitoring.coreos.com_servicemonitors.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.47.0/example/prometheus-operator-crd/monitoring.coreos.com_thanosrulers.yaml

From 13.x to 14.x

Version 14 upgrades prometheus-operator from 0.45.x to 0.46.x. Helm does not automatically upgrade or install new CRDs on a chart upgrade, so you have to install the CRDs manually before updating:

kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.46.0/example/prometheus-operator-crd/monitoring.coreos.com_alertmanagerconfigs.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.46.0/example/prometheus-operator-crd/monitoring.coreos.com_alertmanagers.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.46.0/example/prometheus-operator-crd/monitoring.coreos.com_podmonitors.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.46.0/example/prometheus-operator-crd/monitoring.coreos.com_probes.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.46.0/example/prometheus-operator-crd/monitoring.coreos.com_prometheuses.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.46.0/example/prometheus-operator-crd/monitoring.coreos.com_servicemonitors.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.46.0/example/prometheus-operator-crd/monitoring.coreos.com_thanosrulers.yaml

From 12.x to 13.x

Version 13 upgrades prometheus-operator from 0.44.x to 0.45.x. Helm does not automatically upgrade or install new CRDs on a chart upgrade, so you have to install the CRD manually before updating:

kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.45.0/example/prometheus-operator-crd/monitoring.coreos.com_alertmanagerconfigs.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.45.0/example/prometheus-operator-crd/monitoring.coreos.com_prometheuses.yaml
kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/v0.45.0/example/prometheus-operator-crd/monitoring.coreos.com_alertmanagers.yaml

From 11.x to 12.x

Version 12 upgrades prometheus-operator from 0.43.x to 0.44.x. Helm does not automatically upgrade or install new CRDs on a chart upgrade, so you have to install the CRD manually before updating:

kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/release-0.44/example/prometheus-operator-crd/monitoring.coreos.com_prometheuses.yaml

The chart was migrated to support only helm v3 and later.

From 10.x to 11.x

Version 11 upgrades prometheus-operator from 0.42.x to 0.43.x. Starting with 0.43.x an additional AlertmanagerConfigs CRD is introduced. Helm does not automatically upgrade or install new CRDs on a chart upgrade, so you have to install the CRD manually before updating:

kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/release-0.43/example/prometheus-operator-crd/monitoring.coreos.com_alertmanagerconfigs.yaml

Version 11 removes the deprecated tlsProxy via ghostunnel in favor of native TLS support the prometheus-operator gained with v0.39.0.

From 9.x to 10.x

Version 10 upgrades prometheus-operator from 0.38.x to 0.42.x. Starting with 0.40.x an additional Probes CRD is introduced. Helm does not automatically upgrade or install new CRDs on a chart upgrade, so you have to install the CRD manually before updating:

kubectl apply -f https://raw.githubusercontent.com/prometheus-operator/prometheus-operator/release-0.42/example/prometheus-operator-crd/monitoring.coreos.com_probes.yaml

From 8.x to 9.x

Version 9 of the helm chart removes the existing additionalScrapeConfigsExternal in favour of additionalScrapeConfigsSecret. This change lets users specify the secret name and secret key to use for the additional scrape configuration of prometheus. This is useful for users that have prometheus-operator as a subchart and also have a template that creates the additional scrape configuration.

From 7.x to 8.x

Due to new template functions being used in the rules in version 8.x.x of the chart, an upgrade to Prometheus Operator and Prometheus is necessary in order to support them. First, upgrade to the latest version of 7.x.x

helm upgrade [RELEASE_NAME] prometheus-community/kube-prometheus-stack --version 7.5.0

Then upgrade to 8.x.x

helm upgrade [RELEASE_NAME] prometheus-community/kube-prometheus-stack --version [8.x.x]

Minimal recommended Prometheus version for this chart release is 2.12.x

From 6.x to 7.x

Due to a change in grafana subchart, version 7.x.x now requires Helm >= 2.12.0.

From 5.x to 6.x

Due to a change in deployment labels of kube-state-metrics, the upgrade requires helm upgrade --force in order to re-create the deployment. If this is not done an error will occur indicating that the deployment cannot be modified:

invalid: spec.selector: Invalid value: v1.LabelSelector{MatchLabels:map[string]string{"app.kubernetes.io/name":"kube-state-metrics"}, MatchExpressions:[]v1.LabelSelectorRequirement(nil)}: field is immutable

If this error has already been encountered, a helm history command can be used to determine which release has worked, then helm rollback to the release, then helm upgrade --force to this new one

Configuration

See Customizing the Chart Before Installing. To see all configurable options with detailed comments:

helm show values prometheus-community/kube-prometheus-stack

You may also helm show values on this chart's dependencies for additional options.

Multiple releases

The same chart can be used to run multiple Prometheus instances in the same cluster if required. To achieve this, it is necessary to run only one instance of prometheus-operator and a pair of alertmanager pods for an HA configuration, while all other components need to be disabled. To disable a dependency during installation, set kubeStateMetrics.enabled, nodeExporter.enabled and grafana.enabled to false.

Work-Arounds for Known Issues

Running on private GKE clusters

When Google configure the control plane for private clusters, they automatically configure VPC peering between your Kubernetes cluster’s network and a separate Google managed project. In order to restrict what Google are able to access within your cluster, the firewall rules configured restrict access to your Kubernetes pods. This means that in order to use the webhook component with a GKE private cluster, you must configure an additional firewall rule to allow the GKE control plane access to your webhook pod.

You can read more information on how to add firewall rules for the GKE control plane nodes in the GKE docs

Alternatively, you can disable the hooks by setting prometheusOperator.admissionWebhooks.enabled=false.

PrometheusRules Admission Webhooks

With Prometheus Operator version 0.30+, the core Prometheus Operator pod exposes an endpoint that will integrate with the validatingwebhookconfiguration Kubernetes feature to prevent malformed rules from being added to the cluster.

How the Chart Configures the Hooks

A validating and mutating webhook configuration requires the endpoint to which the request is sent to use TLS. It is possible to set up custom certificates to do this, but in most cases, a self-signed certificate is enough. The setup of this component requires some more complex orchestration when using helm. The steps are created to be idempotent and to allow turning the feature on and off without running into helm quirks.

A pre-install hook provisions a certificate into the same namespace using a format compatible with provisioning using end-user certificates. If the certificate already exists, the hook exits.
The prometheus operator pod is configured to use a TLS proxy container, which will load that certificate.
Validating and Mutating webhook configurations are created in the cluster, with their failure mode set to Ignore. This allows rules to be created by the same chart at the same time, even though the webhook has not yet been fully set up - it does not have the correct CA field set.
A post-install hook reads the CA from the secret created by step 1 and patches the Validating and Mutating webhook configurations. This process will allow a custom CA provisioned by some other process to also be patched into the webhook configurations. The chosen failure policy is also patched into the webhook configurations

Alternatives

It should be possible to use jetstack/cert-manager if a more complete solution is required, but it has not been tested.

You can enable automatic self-signed TLS certificate provisioning via cert-manager by setting the prometheusOperator.admissionWebhooks.certManager.enabled value to true.

Limitations

Because the operator can only run as a single pod, there is potential for this component failure to cause rule deployment failure. Because this risk is outweighed by the benefit of having validation, the feature is enabled by default.

Developing Prometheus Rules and Grafana Dashboards

This chart Grafana Dashboards and Prometheus Rules are just a copy from prometheus-operator/prometheus-operator and other sources, synced (with alterations) by scripts in hack folder. In order to introduce any changes you need to first add them to the original repo and then sync there by scripts.

Further Information

For more in-depth documentation of configuration options meanings, please see

prometheus.io/scrape

The prometheus operator does not support annotation-based discovery of services, using the PodMonitor or ServiceMonitor CRD in its place as they provide far more configuration options. For information on how to use PodMonitors/ServiceMonitors, please see the documentation on the prometheus-operator/prometheus-operator documentation here:

By default, Prometheus discovers PodMonitors and ServiceMonitors within its namespace, that are labeled with the same release tag as the prometheus-operator release. Sometimes, you may need to discover custom PodMonitors/ServiceMonitors, for example used to scrape data from third-party applications. An easy way of doing this, without compromising the default PodMonitors/ServiceMonitors discovery, is allowing Prometheus to discover all PodMonitors/ServiceMonitors within its namespace, without applying label filtering. To do so, you can set prometheus.prometheusSpec.podMonitorSelectorNilUsesHelmValues and prometheus.prometheusSpec.serviceMonitorSelectorNilUsesHelmValues to false.

Migrating from stable/prometheus-operator chart

Zero downtime

Since kube-prometheus-stack is fully compatible with the stable/prometheus-operator chart, a migration without downtime can be achieved. However, the old name prefix needs to be kept. If you want the new name please follow the step by step guide below (with downtime).

You can override the name to achieve this:

helm upgrade prometheus-operator prometheus-community/kube-prometheus-stack -n monitoring --reuse-values --set nameOverride=prometheus-operator

Note: It is recommended to run this first with --dry-run --debug.

Redeploy with new name (downtime)

If the prometheus-operator values are compatible with the new kube-prometheus-stack chart, please follow the below steps for migration:

The guide presumes that chart is deployed in monitoring namespace and the deployments are running there. If in other namespace, please replace the monitoring to the deployed namespace.

Patch the PersistenceVolume created/used by the prometheus-operator chart to Retain claim policy:
```
kubectl patch pv/<PersistentVolume name> -p '{"spec":{"persistentVolumeReclaimPolicy":"Retain"}}'
```
Note: To execute the above command, the user must have a cluster wide permission. Please refer Kubernetes RBAC
Uninstall the prometheus-operator release and delete the existing PersistentVolumeClaim, and verify PV become Released.
```
helm uninstall prometheus-operator -n monitoring
kubectl delete pvc/<PersistenceVolumeClaim name> -n monitoring
```
Additionally, you have to manually remove the remaining prometheus-operator-kubelet service.
```
kubectl delete service/prometheus-operator-kubelet -n kube-system
```
You can choose to remove all your existing CRDs (ServiceMonitors, Podmonitors, etc.) if you want to.

Remove current spec.claimRef values to change the PV's status from Released to Available.

kubectl patch pv/<PersistentVolume name> --type json -p='[{"op": "remove", "path": "/spec/claimRef"}]' -n monitoring

Note: To execute the above command, the user must have a cluster wide permission. Please refer to Kubernetes RBAC

After these steps, proceed to a fresh kube-prometheus-stack installation and make sure the current release of kube-prometheus-stack matching the volumeClaimTemplate values in the values.yaml.

The binding is done via matching a specific amount of storage requested and with certain access modes.

For example, if you had storage specified as this with prometheus-operator:

volumeClaimTemplate:
  spec:
    storageClassName: gp2
    accessModes: ["ReadWriteOnce"]
    resources:
     requests:
       storage: 50Gi

You have to specify matching volumeClaimTemplate with 50Gi storage and ReadWriteOnce access mode.

Additionally, you should check the current AZ of your legacy installation's PV, and configure the fresh release to use the same AZ as the old one. If the pods are in a different AZ than the PV, the release will fail to bind the existing one, hence creating a new PV.

This can be achieved either by specifying the labels through values.yaml, e.g. setting prometheus.prometheusSpec.nodeSelector to:

nodeSelector:
  failure-domain.beta.kubernetes.io/zone: east-west-1a

or passing these values as --set overrides during installation.

The new release should now re-attach your previously released PV with its content.

Migrating from coreos/prometheus-operator chart

The multiple charts have been combined into a single chart that installs prometheus operator, prometheus, alertmanager, grafana as well as the multitude of exporters necessary to monitor a cluster.

There is no simple and direct migration path between the charts as the changes are extensive and intended to make the chart easier to support.

The capabilities of the old chart are all available in the new chart, including the ability to run multiple prometheus instances on a single cluster - you will need to disable the parts of the chart you do not wish to deploy.

You can check out the tickets for this change here and here.

High-level overview of Changes

Added dependencies

The chart has added 3 dependencies.

Node-Exporter, Kube-State-Metrics: These components are loaded as dependencies into the chart, and are relatively simple components
Grafana: The Grafana chart is more feature-rich than this chart - it contains a sidecar that is able to load data sources and dashboards from configmaps deployed into the same cluster. For more information check out the documentation for the chart

Kubelet Service

Because the kubelet service has a new name in the chart, make sure to clean up the old kubelet service in the kube-system namespace to prevent counting container metrics twice.

Persistent Volumes

If you would like to keep the data of the current persistent volumes, it should be possible to attach existing volumes to new PVCs and PVs that are created using the conventions in the new chart. For example, in order to use an existing Azure disk for a helm release called prometheus-migration the following resources can be created:

apiVersion: v1
kind: PersistentVolume
metadata:
  name: pvc-prometheus-migration-prometheus-0
spec:
  accessModes:
  - ReadWriteOnce
  azureDisk:
    cachingMode: None
    diskName: pvc-prometheus-migration-prometheus-0
    diskURI: /subscriptions/f5125d82-2622-4c50-8d25-3f7ba3e9ac4b/resourceGroups/sample-migration-resource-group/providers/Microsoft.Compute/disks/pvc-prometheus-migration-prometheus-0
    fsType: ""
    kind: Managed
    readOnly: false
  capacity:
    storage: 1Gi
  persistentVolumeReclaimPolicy: Delete
  storageClassName: prometheus
  volumeMode: Filesystem

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  labels:
    app.kubernetes.io/name: prometheus
    prometheus: prometheus-migration-prometheus
  name: prometheus-prometheus-migration-prometheus-db-prometheus-prometheus-migration-prometheus-0
  namespace: monitoring
spec:
  accessModes:
  - ReadWriteOnce
  resources:
    requests:
      storage: 1Gi
  storageClassName: prometheus
  volumeMode: Filesystem
  volumeName: pvc-prometheus-migration-prometheus-0

The PVC will take ownership of the PV and when you create a release using a persistent volume claim template it will use the existing PVCs as they match the naming convention used by the chart. For other cloud providers similar approaches can be used.

KubeProxy

The metrics bind address of kube-proxy is default to 127.0.0.1:10249 that prometheus instances cannot access to. You should expose metrics by changing metricsBindAddress field value to 0.0.0.0:10249 if you want to collect them.

Depending on the cluster, the relevant part config.conf will be in ConfigMap kube-system/kube-proxy or kube-system/kube-proxy-config. For example:

kubectl -n kube-system edit cm kube-proxy

apiVersion: v1
data:
  config.conf: |-
    apiVersion: kubeproxy.config.k8s.io/v1alpha1
    kind: KubeProxyConfiguration
    # ...
    # metricsBindAddress: 127.0.0.1:10249
    metricsBindAddress: 0.0.0.0:10249
    # ...    
  kubeconfig.conf: |-
    # ...    
kind: ConfigMap
metadata:
  labels:
    app: kube-proxy
  name: kube-proxy
  namespace: kube-system

24 KiB Raw Blame History Unescape Escape