This section describes how to calculate resources (memory and CPU) required to run Eclipse Che.

Che architectural components

As illustrated in the High-level Che architecture article, Eclipse Che components are:

  • A central workspace controller: an always running service that manages users workspaces

  • Users workspaces: container-based IDEs that the controller stops when the user stops coding

che high level

Both the Che central controller and user workspaces consist of a set of containers. Those containers contribute to the resources consumption in terms of CPU and RAM limits and requests. For a detailed explanation of how Kubernetes manages container-resource requests and limits, see Kubernetes documentation.

Controller requirements

The Workspace Controller consists of a set of five services running in five distinct containers. The following table presents the default resource requirements of each of these services.

Table 1. ControllerServices
Pod Container name Default memory limit Default memory request

Che

che

1 GiB

512 MiB

PostgreSQL

postgres

1 GiB

512 MiB

Keycloak

keycloak

2 GiB

512 MiB

Devfile registry

che-devfile-registry

256 MiB

16 MiB

Plug-in registry

che-plugin-registry

256 MiB

16 MiB

These default values are sufficient when the Che Workspace Controller manages a small amount of Che workspaces. For larger deployments, increase the memory limit. See the Advanced configurations options article for instructions on how to override the default requests and limits. For example, the hosted version of Che that runs on che.openshift.io uses 4 GiB.

Workspaces requirements

This section describes how to calculate the resources required for a workspace. It is the sum of the resources required for each component of this workspace.

These examples demonstrate the necessity of a proper calculation:

  • A workspace with 10 active plug-ins requires more resources then the same workspace with fewer plug-ins.

  • A standard Java workspace requires more resources than a standard Node.js workspace because running builds, tests, and application debugging requires more resources.

Procedure

  1. Identify the workspace components explicitly specified in the components section of the devfile.

  2. Identify the implicit workspace components:

    1. Che implicitly loads the default cheEditor: che-theia, and the chePlugin that allows commands execution: che-machine-exec-plugin. To change the default editor, add a cheEditor component section in the devfile.

    2. When Che is running in multi-user mode, it loads the JWT Proxy component. The JWT Proxy is responsible for the authentication and authorization of the external communications of the workspace components.

  3. Calculate the requirements for each component:

    1. Default values:

      The following table presents the default requirements for all workspace components. It also presents the corresponding Che server property to modify the defaults cluster-wide.

      Table 2. Default requirements of workspace components by type
      Component types Che server property Default memory limit Default memory request

      chePlugin

      che.workspace.sidecar.default_memory_limit_mb

      128 MiB

      128 MiB

      cheEditor

      che.workspace.sidecar.default_memory_limit_mb

      128 MiB

      128 MiB

      kubernetes, openshift, dockerimage

      che.workspace.default_memory_limit_mb, che.workspace.default_memory_request_mb

      1 Gi

      512 MiB

      JWT Proxy

      che.server.secure_exposer.jwtproxy.memory_limit

      128 MiB

      128 MiB

    2. Custom requirements for chePlugins and cheEditors components:

      1. Custom memory limit and request:

        If present, the memoryLimit attribute of the containers section of the meta.yaml file defines the memory limit of a chePlugins or cheEditors component. Che automatically sets the memory request to match the memory limit. There is no attribute in a meta.yaml file to set the memory request.

        Example 1. The chePlugin che-incubator/typescript/latest
        meta.yaml spec section:
        spec:
 containers:
   - image: docker.io/eclipse/che-remote-plugin-node:next
     name: vscode-typescript
     memoryLimit: 512Mi

        It results in a container with the following memory limit and request:

        Memory limit

        512 MiB

        Memory request

        512 MiB
        How to find the meta.yaml file of chePlugin

        Community plug-ins are available in the che-plugin-registry GitHub repository in folder v3/plugins/${organization}/${name}/${version}/.

        For non-community or customized plug-ins, the meta.yaml files are available on the local Kubernetes cluster at ${pluginRegistryEndpoint}/v3/plugins/${organization}/${name}/${version}/meta.yaml.

        For example, on a local Minikube cluster, the URL for the che-incubator/typescript/latest meta.yaml is http://plugin-registry-che.192.168.64.78.nip.io/v3/plugins/che-incubator/typescript/latest/meta.yaml.

      2. Custom CPU limit and request:

        Che does not set CPU limits and requests. The LimitRange Kubernetes namespace may specify some defaults for CPU limits and requests. Kubernetes apply them to the chePlugin container.

    3. Custom requirements for dockerimage components

      If present, the memoryLimit attribute of the devfile defines the memory limit of a dockerimage container. Che set automatically the memory request to match the memory limit. There is no attribute in a devfile to set the memory request.

        - alias: maven
    type: dockerimage
    image: eclipse/maven-jdk8:latest
    memoryLimit: 1536M

    4. Custom requirements for kubernetes or openshift components:

      The referenced manifest may define the memory requirements and limits.

  4. Add all requirements previously calculated.

A workspace example

This section describes a Che workspace example.

The following devfile defines the Che workspace:

apiVersion: 1.0.0
metadata:
 generateName: guestbook-nodejs-sample-
projects:
 - name: guestbook-nodejs-sample
   source:
     type: git
     location: "https://github.com/l0rd/nodejs-sample"
components:
 - type: chePlugin
   id: che-incubator/typescript/latest
 - type: kubernetes
   alias: guestbook-frontend
   reference: https://raw.githubusercontent.com/l0rd/nodejs-sample/master/kubernetes-manifests/guestbook-frontend.deployment.yaml
   mountSources: true
   entrypoints:
     - command: ['sleep']
       args: ['infinity']

This table provides the memory requirements for each workspace component:

Table 3. Total workspace memory requirement and limit
Pod Container name Default memory limit Default memory request

Workspace

theia-ide (default cheEditor)

512 MiB

512 MiB

Workspace

machine-exec (default chePlugin)

128 MiB

128 MiB

Workspace

vscode-typescript (chePlugin)

512 MiB

512 MiB

Workspace

frontend (kubernetes)

1 GiB

512 MiB

JWT Proxy

verifier

128 MiB

128 MiB

Total

2.25 GiB

1.75 GiB

  • The theia-ide and machine-exec components are implicitly added to the workspace, even when not included in the devfile.

  • The resources required by machine-exec are the default for chePlugin.

  • The resources for theia-ide are specifically set in the cheEditor meta.yaml to 512 MiB as memoryLimit.

  • The Typescript VS Code extension has also overridden the default memory limits. In its meta.yaml file, the limits are explicitly specified to 512 MiB.

  • Che is applying the defaults for the kubernetes component type: a memory limit of 1 GiB and a memory request of 512 MiB. This is because the kubernetes component references a Deployment manifest that has a container specification with no resource limits or requests.

  • The JWT container requires 128 MiB of memory.

Adding all together results in 1.75 GiB of memory requests with a 2.25 GiB limit.

Additional resources

Tags: