Base resource requests on actual usage

[NimJay]

Background

• This fixes Base CPU/Memory Requests on Actual Usage #591.
• I looked at a deployment of Online Boutique from 4 days ago. I noted the max (actually used) vCPU and memory values of each microservice — within the full 4 days. (The deployment included the loadgenerator.) Olivier did a similar analysis in 2022.
• For each value, I rounded up to the next 10m (for CPU) and the next 10MiB for memory — to add a buffer.

Change Summary

• I've updated the resource.requests.cpu and resource.requests.memory values of each default microservices's Deployment.

Testing Procedure

• We just have to make sure the staging URL works fine. :)

[github-actions]

🚲 PR staged at http://34.42.127.234

[github-actions]

🚲 PR staged at http://34.42.127.234

[NimJay]

I have tested the new Kubernetes manifests on the regular channel and rapid channel (see delete-me-test-resources-new and delete-me-test-resources-new-rapid below).
The new manifests work.

However, I believe, on GKE autopilot, this only shaved off 0.25 vCPUs because the Managed Service for Prometheus is enabled by default on GKE autopilot and requires a lot (relative to Online Boutique) of resources.

[github-actions]

🚲 PR staged at http://34.42.127.234