Extend pre-flight check to validate container registry access before submitting a run
planned
Rike Hanssen
Runs can fail after submission when the configured credentials allow access to the compute environment but the service account cannot access a private container registry. Today, that access problem may only become clear once the workflow tries to pull an image.
Please explore a preflight check that verifies access to the container registry used by a run before submission. If access is missing, show an actionable error that identifies the registry and explains what needs to be configured. Checking every image in advance may be technically difficult, so the implementation could start by validating registry access where feasible.
M
Michael Tansini
Merged in a post:
Extend Pre-flight checks to check if queues are invalid
Rike Hanssen
A compute environment can still appear available in Platform after its AWS Batch head or worker queue has been deleted or become invalid in AWS. Platform may then accept a run that fails later with a “JobQueue not found” error.
It would be helpful to have a check added the validates that the configured Batch queues exist and are valid when a run is submitted. If a queue is missing or invalid, show an actionable error before the run starts and identify which queue needs attention. This should cover the head and worker queues independently, since either one may be unavailable.
M
Michael Tansini
updated the status to
planned
Planned as part of Q1 27 pre-flight work. The plan is to validate registry access, not container access specifically