Skip to main content
← Back

February 19, 2025

Agenda​

  • DISCUSSION: Combining wash cli and wash lib
  • DISCUSSION: Upgrades and updates for wasmCloud applications

Meeting Notes​

DISCUSSION: Combining wash cli and wash lib​

  • Thanks to ahmedtadde for the work on this initiative.
  • Contributors should hold off on adding major features until this refactor is complete. (Bug fixes are welcome!)
  • There are a few CI issues on the PR at the moment. Maintainers are taking a look at the moment!

DISCUSSION: Upgrades and updates for wasmCloud applications​

  • Today we have a control interface operation to upgrade a component from, say, an OCI reference v1 to an OCI reference v2. The host will verify the component and as soon as it is ready, it'll be hot-swapped in, allowing for a zero-downtime upgrade. We don't do this in wadm or wash dev but the option is available.
  • This can get complicated when you're adding capabilities, changing configuration, etc. This existing approach doesn't seamlessly update links and configs and all of that—you have to do it manually.
  • So we want to have a smoother rollout and rollback story. Brooks has a fairly large RFC that he's put in the GitHub repo: https://github.com/wasmCloud/wasmCloud/issues/4141
  • The purpose here is to introduce robust transition from one version (or configuration) to another and back.
  • For components and providers, it would helpful to report the status at all times.
  • Step 1 is an operator has more information on what the host is doing.
  • Step 2 is to change the way we handle lattice-wide state to have lattice state NATS KV operations managed by wadm or an operator in order to more effectively coordinate operations from multiple hosts.
  • See the diagram on the RFC for a step-by-step walkthrough of how rollouts/rollbacks will work under this plan.
  • Question: A reinitiation of scalers wasn't included in the diagram - is that intentional?
    • Answer: No, good call-out, that should be included in the diagram as well.
  • Question: Are canary deployments or blue-green deployments in scope for this RFC or kept for later?
    • Answer: Probably mostly kept for later. Changing configs can be hard because providers contend for host resources; after introducing the rollouts and rollbacks, a solution here might be to introduce this style of upgrade while excepting the cases where you might have host resource contention.
  • Question: Shouldn't this be an add-on? We've had discussions about an extension model.
    • Answer: This gives the primitives for someone to implement the rolling update at another level.

Recording​