Closed denkv closed 5 years ago
It solves the issue for now. However, I thought that the experiment status is given to the container manager and the pulled images are added to the experiment status object. This would ensure that the set of pulled images does not have to be managed by the container manager since it belongs to the single experiments. :thinking:
the pulled images are added to the experiment status object
This idea came later, but we can implement it now as well.
I would suggest it. It creates some additional effort but I think it is the cleaner implementation.
Now we have a list of image names used during the experiment in ExperimentStatus
.
This will speed up benchmarks like LDCBench which may want to spawn a lot of similar containers.
Fixes #429.