Scalability consideration for abusing reports

[kyujin-cho]

Main idea

This is a (potential) follow-up issue for #737.
The downside of current implementation is that manager need to perform RPC call for each container. When number of container rises, this can cause a performance issue to both manager and agent. We need to fix this.
Some possible options are:

• make agent scan abusing report path periodically, and update redis (or any other state manager) if new container is found
• same, but use heartbeat as a report channel

Any future ideas are welcome.

Alternative ideas

No response

Anything else?

No response