I have the following setup for an AWS Batch Job Queue
const queue = new batch.JobQueue(this, 'Ec2JobQueue', {
computeEnvironments: [
{
computeEnvironment: ec2SpotEnvironment,
order: 1,
},
{
computeEnvironment: ec2OnDemandEnvironment,
order: 2,
},
],
})
The intent being that when i submit a job it will try to provision an ec2 instance by trying the spot market but if that is not possible it will fallback to using on demand.
What i have observed is that it never uses the on demand compute environment. If there are jobs in the queue and I have hit the service quota limit for number of spot instances I can provision then those jobs will just sit in the queue until a spot instance becomes available. I've confirmed that both compute environments are healthy/valid/enabled in the console.
What should be my expectation around when the second compute environment is used? Ideally there would be an additional 'timeout' setting on a compute environment to facilitate something along the lines of (try spot for for 15mins if no luck, try the next compute environment).