Adapt obi-one to use launch-system with AWS-only api/orchestrator deployment#612
Adapt obi-one to use launch-system with AWS-only api/orchestrator deployment#612eleftherioszisis merged 5 commits intomainfrom
Conversation
c6a813a to
7705ffc
Compare
| cores=1, | ||
| memory=2, | ||
| timelimit="00:10", | ||
| compute_cell="local", |
There was a problem hiding this comment.
I was not expecting "local". May be add that one to the documentation.
There was a problem hiding this comment.
"local" is only defined when the launch-system is deployed locally. Alternatively, we could have "cell_a" or "cell_b" as a default.
Note, that I am making here the assumption that the vlab manager request will always return a compute cell domain, therefore this will be always be overridden.
One case that this would be handy could be if we had an env var to disable the request to the vlab manager when obi-one is executed locally alongside a local deployment of the launch-system.
app/dependencies/compute_cell.py
Outdated
| data = response.json() | ||
| compute_cell = data["data"]["virtual_lab"]["compute_cell"] | ||
| L.info("Resolved subdomain %s", compute_cell) | ||
| L.info("Retrieved comput_cell %s", compute_cell) |
There was a problem hiding this comment.
comput_cell -> compute_cell
There was a problem hiding this comment.
Thanks, fixed.
This PR updates obi-one to work with the launch-system now that both the API and orchestrator run exclusively in AWS (no longer in both AWS and Azure).
Summary:
compute_cellto the launch-system in the resources block.obi-one no longer handles cloud-specific routing. The launch-system/orchestrator now determines the correct infrastructure, simplifying the task launch flow and reducing complexity.