Helm in Practice

Kubernetes API Rate Limiting

Two important environment variables control how Helm interacts with the Kubernetes API server in terms of request rate limiting: HELM_QPS and HELM_BURST_LIMIT.

Helm uses the Kubernetes client-go library, which has a built-in rate limiter to control how often it sends requests to the Kubernetes API. This protects the API server from being overloaded during operations such as install, upgrade, or list.

When a chart contains many CRDs, Helm performs extra discovery calls to understand new API groups and resource types, which increases the number of API requests.

HELM_QPS sets the average number of Kubernetes API requests that Helm is allowed to make per second. It defines the steady, sustainable request rate used by the client-go library inside Helm.

• A higher QPS value speeds up operations that require many API calls.
• A lower QPS value reduces the load on a busy or sensitive cluster.

If the value is set to 0 (the default in our case), Helm uses client-go's built-in default, which is usually 5 requests per second.

The burst limit (HELM_BURST_LIMIT) defines how many API requests Helm can send in a short spike before the rate limiter forces it to slow down