a job-level queuing and orchestration module that admit job to be scheduled by some strategies.
refactor skeleton codes to make control flow cleaner.
Why is this needed:
as for now, kubedl is capable to manage lifecycle of job workloads, manipulate pods from job started to succeeded or failed, however, in production users usually requests jobs(=resources) more resources than cluster capacity, therefore, job s have to wait in queue before getting chance to be schedule, for the reason that we propose a new module&policy to coordinate job schedule orders.
assign job priority via schedulePolicy
coordinator job schedule orders by job-coordinator.
What would you like to be added:
Why is this needed:
as for now, kubedl is capable to manage lifecycle of job workloads, manipulate pods from job started to succeeded or failed, however, in production users usually requests jobs(=resources) more resources than cluster capacity, therefore, job s have to wait in queue before getting chance to be schedule, for the reason that we propose a new module&policy to coordinate job schedule orders.
schedulePolicy
job-coordinator
.