feat: set cpu/mem requests and limits for prisma and cloudsql

[pmespresso]

addresses #271

particularly the exceeds its request of 0 as mentioned #271 (comment)

I had previously mistaken the mem/cpu requests fields to be only relate to HPA (which we agreed was unnecessary for our app), which was why that PR was closed.

This PR sets the requests for resources to be as following:

prisma:

	mem	cpu
request	1.28GiB (failed at 1278848Ki === 1.219GiBi)	250m default (wasn't the cause of failure)
limit	2.56GiB	500m

cloudsql

	mem	cpu
request	16Mi (failed at 9364Ki === 9.145Mi)	250m default (wasn't the cause of failure)
limit	32Mi	500m

see also:

• https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/
• https://medium.com/kubernetes-tutorials/efficient-node-out-of-resource-management-in-kubernetes-67f158da6e59
• https://kubernetes.io/docs/tasks/administer-cluster/out-of-resource/

[Tbaut]

Thank you so much for digging into that, and sharing the resources, I think I understood most of it. One question, where did you get the value for the failing (to then guess the request), and how did you decide the limit?

[pmespresso]

where did you get the value for the failing (to then guess the request)

actually you did :) #271 (comment), i.e. kubectl describe <svc>

[Tbaut]

Hmm, I think I didn't write this clearly. The pod describe was on the last 50k. I don't think the nodewatcher ever got evicted. But I can't remember exactly tbh.

[pmespresso]

hmm but the containers that ran out of resources according to that log are in nodewatcher deployment.