What you can optimize in Che

Identify the performance areas you can tune in Che so that you choose the right optimization strategy for your deployment.

A default Che installation runs all server components with a single replica and no image pre-caching. This configuration works for small teams but can cause slow workspace starts and limited capacity as usage grows. You can optimize two areas independently:

Workspace startup speed

When a developer starts a workspace, Kubernetes pulls container images from the registry. On a cold node, this can take several minutes. The Kubernetes Image Puller pre-caches images on every node so that workspaces start in seconds. For instructions on deploying and configuring the Image Puller, see Additional resources.

Platform scaling

By default, each Che server component runs with one replica. You can increase replicas for high availability and configure cluster autoscaling to add worker nodes when demand exceeds capacity. For instructions on configuring autoscaling, see Additional resources.

When to optimize

Consider optimization when you observe:

  • Workspace startup taking more than 60 seconds on nodes that have not previously run workspaces.

  • Dashboard or gateway timeouts during peak usage when many developers start workspaces simultaneously.

  • Workspace scheduling failures because of insufficient node resources.

What affects workspace startup and platform capacity

Area Control Default

Image pre-caching

CheCluster CR field spec.components.imagePuller.enable

false (disabled)

Server replicas

CheCluster CR field spec.components.<component>.deployment.replicas

1 for each component

Node autoscaling

MachineAutoscaler and ClusterAutoscaler custom resources

Not configured