Large Instances Meeting Notes 2023-10-03

Owned by Braden MacDonald
Last updated: Oct 12, 2023 by Xavier Antoviaque

Meeting recording: https://drive.google.com/file/d/18wel5_Psiv6mHtMQpD8hjfrk1GgvqzM4/view?usp=sharing

Assign meeting lead and note taker.

Greetings & introductions as needed. n/a this time

Updates from each org on the call - 2U, eduNEXT, OpenCraft, Raccoon Gang. What's new with your deployment(s)?

@Jhony Avella : we (eduNEXT) were able to set up an installation with a single-node k8s cluster using k3s, including all the databases (MySQL, Mongo, Redis). We applied a couple of load tests. We were able to support up to 60 concurrent users (w/ 4 cpus, 16 CPU); or up to 30 concurrent in the browser tests.

@Felipe Montoya : We’re experimenting with k3s because we’re looking for something to fill the gap for customers that don’t need a full k8s cluster but need something more scalable and flexible than just a single VM. So far it seems promising. What we’ve seen is that “tutor local” is tuned for very small instances and the moment you get significant traffic, it doesn’t handle it well. So this approach is more attractive for small production instances.

@Moisés González : It also allows us to use the same deployment tools as before beacause it's still kubernetes.

@Jeremy Bowman (Deactivated) : The Arch-BOM team at 2U is getting closer to focusing on dev environments and deployments; we’ve finished dealing with a number of issues that were taking up a lot of our time, so excited to get to those projects. Learning in depth about k8s from the book. Also looking at Orbstack, and it’s really nice, has lots of integrations to let you explore the file system easily. There’s also Colima which is open source (has no UI though). What’s better about Orbstack is that it’s 3x as fast as Colima/DockerDesktop, and has all these extra developer convenience features.

@Gábor Boros : not much to report. @Braden MacDonald : trying to free up some of Gabor’s time so we can work on the Harmony integration into our internal Grove pipeline.

 

Harmony project updates:

  • we’re still blocked on publishing official Helm chart until at least two orgs are using it in production. eduNEXT is testing it on two clusters, now need OpenCraft or RacoonGang to do the same.

  • Karpenter PR: @Jhony Avella I added a couple of resources that are required for Karpenter to work properly as well as a newer version of the Karpenter Helm chart. Problem is that Karpenter needs a couple of additional CRDs after installing the Helm chart, and Helm doesn’t manage CRDs. Trying to install them as part of the Harmony Helm chart was causing a failure so for now you have to install them manually.

    • @Felipe Montoya How important is Karpenter/AWS? Do we have a sense of which providers will be using it and to what extent? Answer: lots of interest but nobody on the call is using it yet.

  • Prometheus/Grafana ticket: we may want to revisit this in light of Aspects; maybe use Superset instead etc.

  • Support for Google Cloud: we discussed closing the issue unless/until someone is interested in using it. Current providers on the call are not using GKE.

 

Open discussion/questions, if any.

@Felipe Montoya : Last week I asked to Frontend Working Group about the switch to Paragon Alpha, which includes design tokens and CSS variables, and I previously thought it was going to be merged in October and released with Quince. But I found out that’s unlikely, and it definitely won’t be in all the MFEs by Quince. We are trying to push it forward via the Frontend WG to get it into Redwood because it will make theming many MFEs much easier.

@Braden MacDonald Related: there is a frontend pluggability summit coming up on Oct. 25 https://discuss.openedx.org/t/frontend-pluggability-summit/11167