[catalyst-latam] Standarize profileList

[AIDEA775]

This changes was deployed in staging hub: https://staging.latam.catalystproject.2i2c.cloud

ref: #3631

[github-actions]

Merging this PR will trigger the following deployment actions.

Support and Staging deployments

Cloud Provider	Cluster Name	Upgrade Support?	Reason for Support Redeploy	Upgrade Staging?	Reason for Staging Redeploy
gcp	catalystproject-latam	No		Yes	Following helm chart values files were modified: common.values.yaml

Production deployments

Cloud Provider	Cluster Name	Hub Name	Reason for Redeploy
gcp	catalystproject-latam	unitefa-conicet	Following helm chart values files were modified: common.values.yaml
gcp	catalystproject-latam	cicada	Following helm chart values files were modified: common.values.yaml, cicada.values.yaml
gcp	catalystproject-latam	gita	Following helm chart values files were modified: common.values.yaml

[consideRatio]

Nice!! I suggest we try to go for a merge and iterate from there!

There are some inefficiencies around requesting 8GB exactly for example, as that would only fit 3 users on a 32 GB node due to some capacity is reserved by k8s and some is needed by a few system pods running on the node besides the user pods. I discussed this in #3631 (comment).

I was when I wrote that more positive towards use of n2-highmem-2 nodes, but I think now that I've remembered the complexities about overhead that becomes relatively large on smaller node makes me think its good to avoid use of a combination of 2 CPU and 4 CPU nodes, and figure it may very well be that 4 CPU nodes is needed to fit a few users per node in order to avoid regular startup delays.

Deliberations

• Should memory requests be lowered to account for 32 GB not being available?
  I think let's not try to do it and make the UX simple, if you request 1, 2, 4, 8GB, you get exactly that. At worst we use ~25% of the capacity inefficiently on a 4 CPU node, which happens if all users start the 8 GB servers.

• Should we have a node running?
  I think having one node running at all time would be a reasonable optimization to ensure startup times. It adds cloud costs, but its quite a big deal I think to get quick access.

• Should n2-highmem-2 or n2-highmem-4 nodes be used?
  Using 2 CPU / 16GB node, the 8GB requests would only fit a single user pod on the node and wasting at worst ~50% of the capacity. If we use 2 CPU nodes, I think at least the 8 GB request must be tweaked to ensure at least 2 such pods would fit by themselves on a node to at worst waste ~25% of the capacity.

  Overall, I lean towards thinking that use of 4 CPU nodes is a better compromise weighing in operational costs etc of handling 2 CPU nodes, that may at some point be considered too small since they could force new node startups too often if users requests 4 GB / 8 GB in classes of 30 for example.

[github-actions]

🎉🎉🎉🎉

Monitor the deployment of the hubs here 👉 https://github.com/2i2c-org/infrastructure/actions/runs/8116989943