[tools][dependency jobs][scheduling] Request to support reasonable START_AFTER on ipmi.
Currently for the dependent jobs with START_AFTER relationship, eg A START_AFTER B, openqa scheduling can ensure that A is after B, but can not promise that:
- A is just the next job of B(possible a job C(os installation) in between)
- A and B run on the same worker
This is fine for jobs on qemu workers. However for ipmi jobs, the above two constraints are needed. For example, a quite common request on ipmi machines is that first a host installation is done, and then launch various kinds of tests. This is a good example that should use START_AFTER relationship from a common tool's view. However due to above limitations, we can not do it on openqa.
With more and more users on ipmi workers to test physical machines, eg sebastian's team , QAM team and sle-virt team, we strongly recommend to add support for the stricter START_AFTER on ipmi machines. Especially currently we kind of partially rely on it to improve the openqa test efficiency to get prepare for future parallel developing products tested the same time on openqa.