Closed clintkitson closed 8 years ago
The plan as discussed is to create a library that abstracts the mesos containerizer. This way the mesos-module-dvdi
can call other isolators on it's own.
In the future, a proper order of events for prepare(), cleanup() etc would likely be desirable to containerizers that call each other.
The current module at version 0.1 is limiting due to the Isolator implementation. Currently, there is only one type of filesystem based Isolator that can be used at a time. This means that if you are using the
mesos-module-dvdi
you cannot isolator the filesystem to a single container, it must be exposed to all containers.The problem is due to the fact that Isolators don't pass mount points between themselves. The mount point is only determined after
mesos-module-dvdi
runs which then needs to be passed to the next Isolator to privately mount the filesystem in the container.