Using InfluxDB for Analytics

Flagsmith has a soft dependency on InfluxDB to store time-series data. You don't need to configure InfluxDB to run the platform; by default, this data will be stored in PostgreSQL. If you are running very high traffic loads, you might be interested in deploying InfluxDB.

Docker

1. Create a user account in InfluxDB. You can visit [http://localhost:8086/]
2. Go into Data > Buckets and create three new buckets called default, default_downsampled_15m and default_downsampled_1h
3. Go into Data > Tokens and grab your access token.
4. Edit the docker-compose.yml file and add the following environment variables in the API service to connect the API to InfluxDB:
   • INFLUXDB_TOKEN: The token from the step above
   • INFLUXDB_URL: http://influxdb
   • INFLUXDB_ORG: The organisation ID - you can find it here
   • INFLUXDB_BUCKET: default
5. Restart docker-compose
6. Create a new task with the following query. This will downsample your per-millisecond API request data down to 15-minute blocks for faster queries. Set it to run every 15 minutes.

option task = {name: "Downsample (API Requests)", every: 15m}

data = from(bucket: "default")
 |> range(start: -duration(v: int(v: task.every) * 2))
 |> filter(fn: (r) =>
  (r._measurement == "api_call"))

data
 |> aggregateWindow(fn: sum, every: 15m)
 |> filter(fn: (r) =>
  (exists r._value))
 |> to(bucket: "default_downsampled_15m")

Once this task has run, you will see data coming into the Organisation API Usage area.

7. Create another new task with the following query. This will downsample your per-millisecond flag evaluation data down to 15-minute blocks for faster queries. Set it to run every 15 minutes.

option task = {name: "Downsample (Flag Evaluations)", every: 15m}

data = from(bucket: "default")
 |> range(start: -duration(v: int(v: task.every) * 2))
 |> filter(fn: (r) =>
  (r._measurement == "feature_evaluation"))

data
 |> aggregateWindow(fn: sum, every: 15m)
 |> filter(fn: (r) =>
  (exists r._value))
 |> to(bucket: "default_downsampled_15m")

Once this task has run, and you have made some flag evaluations with analytics enabled (see documentation here for information on this), you should see data in the 'Analytics' tab against each feature in your dashboard.

8. Create another new task with the following query. This will downsample your per-millisecond API request data down to 1-hour blocks for faster queries. Set it to run every 1 hour.

option task = {name: "Downsample API 1h", every: 1h}

data = from(bucket: "default")
 |> range(start: -duration(v: int(v: task.every) * 2))
 |> filter(fn: (r) =>
  (r._measurement == "api_call"))

data
 |> aggregateWindow(fn: sum, every: 1h)
    |> filter(fn: (r) =>
      (exists r._value))
 |> to(bucket: "default_downsampled_1h")

9. Create another new task with the following query. This will downsample your per-millisecond flag evaluation data down to 1-hour blocks for faster queries. Set it to run every 1 hour.

option task = {name: "Downsample API 1h - Flag Analytics", every: 1h}

data = from(bucket: "default")
 |> range(start: -duration(v: int(v: task.every) * 2))
 |> filter(fn: (r) =>
  (r._measurement == "feature_evaluation"))
 |> filter(fn: (r) =>
  (r._field == "request_count"))
 |> group(columns: ["feature_id", "environment_id"])

data
 |> aggregateWindow(fn: sum, every: 1h)
    |> filter(fn: (r) =>
      (exists r._value))
 |> set(key: "_measurement", value: "feature_evaluation")
 |> set(key: "_field", value: "request_count")
 |> to(bucket: "default_downsampled_1h")

Kubernetes (via Helm)

By default, Flagsmith uses PostgreSQL to store time-series data. You can alternatively use InfluxDB to track:

• SDK API traffic
• SDK Flag Evaluations

You will need to perform the steps above to configure InfluxDB itself. You can then configure the Helm chart with the following values:

Parameter	Description	Default
influxdb2.enabled		true
influxdb2.nameOverride		influxdb
influxdb2.image.repository	Docker image repository for InfluxDB	quay.io/influxdb/influxdb
influxdb2.image.tag	Docker image tag for InfluxDB	v2.0.2
influxdb2.image.imagePullPolicy		IfNotPresent
influxdb2.image.imagePullSecrets		[]
influxdb2.adminUser.organization		influxdata
influxdb2.adminUser.bucket		default
inflxdb2.adminUser.user		admin
influxdb2.adminUser.password		randomly generated
influxdb2.adminUser.token		randomly generated
influxdb2.persistence.enabled		false
influxdb.resources	resources per pod for the InfluxDB	{}
influxdb.nodeSelector		{}
influxdb.tolerations		[]
influxdb.affinity		{}
influxdbExternal.enabled	Use an InfluxDB not managed by this chart	false
influxdbExternal.url
influxdbExternal.bucket
influxdbExternal.organization
influxdbExternal.token
influxdbExternal.tokenFromExistingSecret.enabled	Use reference to a k8s secret not managed by this chart	false
influxdbExternal.tokenFromExistingSecret.name	Referenced secret name
influxdbExternal.tokenFromExistingSecret.key	Key within the referenced secret to use