Issue Type:

• Bug report
• Feature request

What happened:
Clients do not correctly close connections, or network components silently drop connections without closing or resetting them.
After several days this can result in Linkerd running out of file handles.

What you expected to happen:
It should be possible to expire server connections.

How to reproduce it (as minimally and precisely as possible):
Start Linkerd with a very low limit of file handles.
Then open as many connections as possible to Linkerd, can be a plain socket, until Linkerd fails to accept the connections due to lack of file handles.

Anything else we need to know?:

Environment:

• linkerd/namerd version, config files:
• Platform, version, and config files (Kubernetes, DC/OS, etc): No Kubernetes or such, just plain VMs
• Cloud provider or hardware configuration: