GKE Autopilot Tutorial: Resource requirement update for userservice

[oginskis]

The Autopilot sets limits to request values (reference: https://cloud.google.com/kubernetes-engine/docs/concepts/autopilot-resource-requests#resource-limits) therefore the userservice gets only up to 200 milicores of CPU at maximum on the Autopilot. The userservice generates JWT tokens which are CPU-consuming and currently it takes several seconds for a user to log in.

I propose to set the CPU requests to 500m to improve the UX.

@bourgeoisor

[bourgeoisor]

Hi @oginskis! Two quick notes:

1. The kubernetes-manifests/ directory is auto-generated with new releases and should not be manually modified. Instead, could you make your changes in dev-kubernetes-manifests/? Those changes are then picked up during the next release.

2. I tested these changes in a GKE Standard default configuration and wasn't able to deploy successfully (not enough vCPU to deploy). Bank of Anthos currently can be deployed in a default GKE Standard cluster without any changes or scaling, and this needs to stay true for the many quickstart flows it supports. I have tried the following which did work, though: bumping the vCPU by 50% from 200m to 300m, as well as doubling memory from 64Mi to 128Mi. Would that work for you?

Feel free to make those two changes and I'll re-verify this PR

[NimJay]

@bourgeoisor first off, thank you for thoroughly testing this pull-request.

Question: Have I interpreted your suggestions correctly?

[bourgeoisor]

Yep, thank you very much, Nim!

[bourgeoisor]

LGTM