Skip to main content
Enterprise plan Helm chart
This page is for engineering teams self-hosting their own Lightdash instance. If you want to get started with pre-aggregates, see the pre-aggregates reference.
We recommend deploying Lightdash with pre-aggregates using the Helm chart. The Helm chart handles the required service dependencies and environment variable wiring automatically.

​
Enabling pre-aggregates

Pre-aggregates materialize query results so that repeated queries are served from DuckDB instead of hitting your data warehouse. This requires NATS for async job processing and S3-compatible storage for materialized results.

​
Prerequisites

  • A valid Lightdash license key
  • An S3-compatible bucket (AWS S3, GCS, MinIO, etc.)

​
Helm values

Setting these three values in your Helm values is the minimum required configuration: 
# Enable NATS and workers
nats:
  enabled: true
warehouseNatsWorker:
  enabled: true
preAggregateNatsWorker:
  enabled: true

# License key
secrets:
  LIGHTDASH_LICENSE_KEY: "your-license-key"

# S3 storage for materialized results
configMap:
  S3_ENDPOINT: "https://s3.us-east-1.amazonaws.com"
  S3_REGION: "us-east-1"
  PRE_AGGREGATE_RESULTS_S3_BUCKET: "my-lightdash-pre-aggs"

secrets:
  S3_ACCESS_KEY: "your-access-key"
  S3_SECRET_KEY: "your-secret-key"
The chart auto-configures NATS_ENABLED, PRE_AGGREGATES_ENABLED, NATS_URL, and PRE_AGGREGATES_PARQUET_ENABLED from the flags above.

​
What gets deployed

ComponentPurpose
NATS JetStreamMessage broker for async query jobs
Warehouse workerProcesses interactive queries from users
Pre-aggregate workerMaterializes pre-aggregates and processes DuckDB queries
Warehouse and pre-aggregate workers are separate deployments so they don’t compete for resources.

​
Scaling

The defaults are tuned for typical workloads. The main levers if you need to adjust: 
warehouseNatsWorker:
  replicas: 1        # scale horizontally for more concurrent queries
  concurrency: 100   # concurrent jobs per pod

preAggregateNatsWorker:
  replicas: 1
  concurrency: 100
Pre-aggregate workers are more resource-intensive than warehouse workers because they run DuckDB. The default resource requests reflect this:
Warehouse workerPre-aggregate worker
CPU250m650m
Memory1.5Gi4Gi
Ephemeral storage9Gi9Gi