Skip to content

Latest commit

聽

History

History
136 lines (108 loc) 路 7.6 KB

README.md

File metadata and controls

136 lines (108 loc) 路 7.6 KB
Raw

Promscale Helm chart

This directory contains a Helm chart to deploy the Promscale Connector on Kubernetes. If you are looking to deploy an entire observability suite including Prometheus, Promscale, Grafana and other tools, we suggest using The Observability Suite for Kubernetes (tobs).

This chart will do the following:

  • Create a Kubernetes Deployment (by default) with one pod
  • Create a Kubernetes Service exposing access to the Connector pods
    • By default a LoadBalancer, but can be disabled to only a ClusterIP with a configurable port
  • Create a Kubernetes CronJob that deletes the data chunks that fall out of the retention period

Note: We have dropped compatibility with TimescaleDB 1.x. If you would like to use Promscale with TimescaleDB 1.x then use the helm charts from Promscale release 0.6.0.

Prerequisites

For promscale to work correctly it needs a set of data to connect to timescale database. This configuration can be supplied in two ways either by using DB URI or by specifying connection parameters. Options are mutually exclusive and specifying URI takes priority.

Using DB URI

You can use db uri to connect to TimescaleDB. To do so, specify the URI in values.yaml as follows:

connection:
  uri: <TIMESCALE_DB_URI>

Using Connection Parameters

Instead of using db uri, you can specify all parameters necessary for connecting promscale to timescaledb using connection map. Bear in mind that timescale database should exist before starting promscale or at least credentials should be available.

Following are the default configuration values:

connection:
  user: postgres
  password: ""
  host: timescaledb.default.svc.cluster.local
  port: 5432
  sslMode: require
  dbName: timescale

User-Managed Connection Parameter Secret

Note: We primarily provide this feature for workflow convenience, but it is an advanced feature requiring more care to configure correctly. We suggest that you configure the connection options in values.yaml instead. If you are using GitOps and do not want to store database passwords unencrypted in your Git repo, you probably want to use tooling like helm-secrets instead.

If you would like to provide connections details via your own secret which is managed outside the lifecycle of this helm chart, you can do so. The secret must contain the environment variables corresponding to Promscale's connection configuration. Additionally, the connectionSecretName parameter of this helm chart must be configured with the name of your secret, and your secret must be in the same namespace as the deployment.

An example of what this could look like is as follows:

# secret.yaml
apiVersion: v1
kind: Secret
metadata:
  name: my-promscale-connection-secret
stringData:
  PROMSCALE_DB_PORT: "5432"
  PROMSCALE_DB_USER: "postgres"
  PROMSCALE_DB_PASSWORD: "password"
  PROMSCALE_DB_HOST: "db.timescale.svc.cluster.local"
  PROMSCALE_DB_NAME: "timescale"
  PROMSCALE_DB_SSL_MODE: "require"

In the values.yml of the helm chart you would then have the following:

# extract from values.yaml
connectionSecretName: my-promscale-connection-secret

Note: The Promscale pod, once started, will not automatically take on any changes to the contents of connectionSecretName.

Installing

To install the chart with the release name my-release:

helm install --name my-release .

You can override parameters using the --set key=value[,key=value] argument to helm install, e.g. to install the chart with specifying a previously created secret timescale-secret and an existing TimescaleDB instance:

helm install --name my-release . \
      --set connectionSecretName="promscale-connection-secret"

You can also install by referencing the db uri secret created previously:

helm install --name my-release . \
      --set connectionSecretName="promscale-connection-secret"

Alternatively, a YAML file the specifies the values for the parameters can be provided while installing the chart. For example:

helm install --name my-release -f myvalues.yaml .

Configuration

Parameter Description Default
image The image (with tag) to pull timescale/promscale
imagePullPolicy The image pull policy IfNotPresent
replicaCount Number of pods for the connector 1
upgradeStrategy Promscale deployment upgrade strategy, By default set to Recreate as during Promscale upgrade we expect no Promscale to be connected to TimescaleDB Recreate
resources Requests and limits for each of the pods {}
nodeSelector Node labels to use for scheduling {}
tolerations Tolerations to use for scheduling []
affinity PodAffinity and PodAntiAffinity settings {}
extraArgs Additional promscale CLI arguments []
connection.uri DB uri string used for database connection. When not empty it takes priority over other settings in connection map. ""
connection.user Username to connect to TimescaleDB with postgres
connection.password The DB password for user specified in connection.user ""
connection.host Hostname of timescaledb instance db.timescaledb.svc.cluster.local
connection.port Port the db listens to 5432
connection.dbName Database name in TimescaleDB to connect to timescale
connection.sslMode SSL mode for connection require
prometheus.port Port the connector Service accepts prometheus remote_write connections on 9201
prometheus.annotations Annotations to allow prometheus metrics collection. { "prometheus.io/scrape": 'true', "prometheus.io/port": '9201', "prometheus.io/path": '/metrics'}
openTelemetry.enabled Enable OpenTelemetry tracing support in Promscale (in future releases tracing support will be enabled by default) false
openTelemetry.port Port the connector Service will accept otlp connections on 9202
service.type Type of Service to be used ClusterIP
service.annotations Annotations to set to the Service {}
serviceMonitor.enabled Enable creation of serviceMonitor object used by prometheus-operator. prometheus.annotations should be set to {} when using this option. false