Github user qianzhangxa commented on the issue:
https://github.com/apache/mesos/pull/263
I'd like to echo @jdef's comment, we need a clear use case for ip per
nested container. Our current status is, if framework launches multiple task
groups (pods) via a single default executor, all the nested containers of all
these task groups will share the executor's network namespace. This is actually
different from Kubernetes pod where each pod will have its own network
namespace and all the container in a pod will share the same network namespace
so that they can communicated with 127.0.0.1/localhost. IMHO, we should
consider to do something similar with Kubernetes, i.e., each task group will
have its own network namespace rather than each nested container has its own
network namespace unless we have a use case for it.
---
