Add resource limits/requests to the devfiles and plugins in the registries

[l0rd]

Is your enhancement related to a problem? Please describe.

If no limits are set the Che default will be inferred. That may result in workspace not able to be scheduled in the Kubernetes cluster.

Describe the solution you'd like

Look for CPU/memory request and limits for each plugins and explicitely specify them. The same for containers components in devfiles.

Additional context

• Specify che-theia CPU limit (issue / PR)
• Specify jwt-proxy CPU limit (issue)
• Specify che-theia-plugins CPU limits (PR)
• Specify che-machine-exec and remote-runtime-injector CPU limits (PR)
• Specify che-theia-plugins memory requests in che-theia-plugin.yaml (PR)
• Add a che-plugin-registry PR check to avoid editros or plugins without mem/cpu requests and limits (PR)
• Evaluate target namespace ResourceQuota and LimitRange to evaluate and applyg the best mem/cpu requests and limits will be solved in the context of this epic

[che-bot]

Issues go stale after 180 days of inactivity. lifecycle/stale issues rot after an additional 7 days of inactivity and eventually close.

Mark the issue as fresh with /remove-lifecycle stale in a new comment.

If this issue is safe to close now please do so.

Moderators: Add lifecycle/frozen label to avoid stale mode.

[gattytto]

I can relate to this as I get "heap overflow" when trying to run/debug the haskell sample in che osio

[ibuziuk]

@ericwill @l0rd could you please discuss this issue during the prioritization? I believe it would be great to have it in the 7.24.0

[gattytto]

haskell tryout is here including devfile

[l0rd]

Discussed yesterday: @azatsarynnyy should create a specific issue for Theia: profiling and setting of request/limits in meta.yaml. Whereas @ericwill team will start profiling strategic plugins (i.e. those that we deliver on CRW).

[gattytto]

this can be a problem

[vitaliy-guliy]

      cpuLimit: 500m
      cpuRequest: 30m

Those limitations will be applied to all the plugins as the first iteration.
It seems 500 millicores is pretty enough and does not significantly affect the plugins.

[svor]

Some plugins are described in che-theia.yaml:

• che-machine-exec
• remote-runtime-injector

need to add resources limits/requests into them as well

[ibuziuk]

@vitaliy-guliy @ericwill @svor folks, could you please clarify what is missing in this epic? Are we already specify limits & requests for every devfile & plugin in our registries?

[svor]

@ibuziuk We still have 3 subtasks to be done, you can see that in the description

[ibuziuk]

@svor right, there are some bullet points in the description but looks like there are no related issues for sub-tasks. is it planned to be addressed in the next sprint according to the label?

[svor]

@l0rd do you want us to take this epic into the next sprint and finish it?

[ibuziuk]

just for the record, default CPU limits might be applied on workspaces.openshift.com infra in the future. So, we need to make sure that all our devfiles & plugins specify the limits & requests explicitly. Otherwise, the default might be applied, which could lead to the performance issues during the workspace usage.

[l0rd]

@svor yes for "Specify che-theia-plugins memory requests in che-theia-plugin.yaml" and "Add a che-plugin-registry PR check to avoid editors or plugins without mem/cpu requests and limits".

The last one has lower priority and needs a dedicated epic (I will create it).

[svor]

@svor yes for "Specify che-theia-plugins memory requests in che-theia-plugin.yaml" and "Add a che-plugin-registry PR check to avoid editors or plugins without mem/cpu requests and limits".

OK, we'll take it into the next sprint

[sunix]

@vitaliy-guliy could we adjust the memory limit for the machine exec plugin?

[vitaliy-guliy]

@vitaliy-guliy could we adjust the memory limit for the machine exec plugin?

Sure, I will try to to give it 128Mb of RAM, as we already set that value for eclipse/che-machine-exec container here https://github.com/eclipse-che/che-plugin-registry/blob/main/che-editors.yaml#L115

[svor]

@svor yes for "Specify che-theia-plugins memory requests in che-theia-plugin.yaml" and "Add a che-plugin-registry PR check to avoid editors or plugins without mem/cpu requests and limits".

The last one has lower priority and needs a dedicated epic (I will create it).

@l0rd "Specify che-theia-plugins memory requests in che-theia-plugin.yaml" and "Add a che-plugin-registry PR check to avoid editors or plugins without mem/cpu requests and limits" subtasks are done.

[l0rd]

@svor I would like to close this epic and open a different one dedicated to the last subtask "Evaluate target namespace ResourceQuota and LimitRange to evaluate and applyg the best mem/cpu requests and limits" with a label roadmap/6-months. What do you think?

[svor]

@l0rd sounds good to me, that was the plan:

@svor yes for "Specify che-theia-plugins memory requests in che-theia-plugin.yaml" and "Add a che-plugin-registry PR check to avoid editors or plugins without mem/cpu requests and limits".

The last one has lower priority and needs a dedicated epic (I will create it).

[l0rd]

Follow up epic for CPU/MEM check at runtime #20509