AWS Batch JobQueue does not provide computeResources required for managed ComputeEnvironment. Since JobQueue does not require the user to provide a computeEnvironmentOrder, the default ComputeEnvironment created by it should provide some computeResources as required by ComputeEnvironment.
Need to be fixed after the change in #6549
JobQueueYou will get the following error
computeResources is missing but required on a managed compute environment
This is :bug: Bug Report
Hmm.. I see no solution other than creating an "unmanaged" computeEnvironment on the JobQueue constructor or make it a required property. Actually I see the latter case as more appropriate.
Defining default computeResources makes no sense to me.
Thoughts?
Yes I agree. Default computeResource is very rare and would only be used when the compute defaults actually match to what you would need (but maybe this is actually pretty common?). I think we can make computeEnvironments a required prop to the queue.
If not, we can accept vpc as a prop and create a default computeEnvironement that the ComputeEnvironment constructor creates (that also just accepts a vpc and nothing else as required props)
I agree that making the computeEnvironments property required at this stage is better. It makes more sense to me for any use case that isn't just sanity testing. Maybe we can think of a default compute environment later on. And even then I would probably prefer having something like ComputeEnvironment.lotsOfMemory() to help users, rather than implicitly creating the defaults.
@andrestone Feel like whipping up a PR?
@ayush987goyal You're right. It was an oversight in the initial PR that introduced these L2's.
@andrestone Has already added some content to the page with the launch template support PR.
I wouldn't want to overwhelm this PR with adding all this content right now. I will create a separate issue for completing the README with all the necessary examples and explanations.
Sound good?