Add support for metrics plugins

Allows for a plugin type that can be used to scrape metrics.
This is useful because metrics are not neccessarily at a standard
location... `--metrics-addr` must be set, and must currently be a TCP
socket.
Even if metrics are done via a unix socket, there's no guarentee where
the socket may be located on the system, making bind-mounting such a
socket into a container difficult (and racey, failure-prone on daemon
restart).

Metrics plugins side-step this issue by always listening on a unix
socket and then bind-mounting that into a known path in the plugin
container.

Note there has been similar work in the past (and ultimately punted at
the time) for consistent access to the Docker API from within a
container.

Why not add metrics to the Docker API and just provide a plugin with
access to the Docker API? Certainly this can be useful, but gives a lot
of control/access to a plugin that may only need the metrics. We can
look at supporting API plugins separately for this reason.

Signed-off-by: Brian Goff <cpuguy83@gmail.com>

docs/extend/config.md

@@ -61,6 +61,8 @@ Config provides the base accessible fields for working with V0 plugin format
 
         - **docker.logdriver/1.0**
 
+        - **docker.metricscollector/1.0**
+
     - **`socket`** *string*
 
       socket is the name of the socket the engine should use to communicate with the plugins.

docs/extend/plugins_metrics.md

@@ -0,0 +1,85 @@
+---
+title: "Docker metrics collector plugins"
+description: "Metrics plugins."
+keywords: "Examples, Usage, plugins, docker, documentation, user guide, metrics"
+---
+
+<!-- This file is maintained within the docker/docker Github
+     repository at https://github.com/docker/docker/. Make all
+     pull requests against that repo. If you see this file in
+     another repository, consider it read-only there, as it will
+     periodically be overwritten by the definitive file. Pull
+     requests which include edits to this file in other repositories
+     will be rejected.
+-->
+
+# Metrics Collector Plugins
+
+Docker exposes internal metrics based on the prometheus format. Metrics plugins
+enable accessing these metrics in a consistent way by providing a Unix
+socket at a predefined path where the plugin can scrape the metrics.
+
+> **Note**: that while the plugin interface for metrics is non-experimental, the naming
+of the metrics and metric labels is still considered experimental and may change
+in a future version.
+
+## Creating a metrics plugin
+
+You must currently set `PropagatedMount` in the plugin `config.json` to
+`/run/docker`. This allows the plugin to receive updated mounts
+(the bind-mounted socket) from Docker after the plugin is already configured.
+
+## MetricsCollector protocol
+
+Metrics plugins must register as implementing the`MetricsCollector` interface
+in `config.json`.
+
+On Unix platforms, the socket is located at `/run/docker/metrics.sock` in the
+plugin's rootfs.
+
+`MetricsCollector` must implement two endpoints:
+
+### `MetricsCollector.StartMetrics`
+
+Signals to the plugin that the metrics socket is now available for scraping
+
+**Request**
+```json
+{}
+```
+
+The request has no playload.
+
+**Response**
+```json
+{
+	"Err": ""
+}
+```
+
+If an error occurred during this request, add an error message to the `Err` field
+in the response. If no error then you can either send an empty response (`{}`)
+or an empty value for the `Err` field. Errors will only be logged.
+
+### `MetricsCollector.StopMetrics`
+
+Signals to the plugin that the metrics socket is no longer available.
+This may happen when the daemon is shutting down.
+
+**Request**
+```json
+{}
+```
+
+The request has no playload.
+
+**Response**
+```json
+{
+	"Err": ""
+}
+```
+
+If an error occurred during this request, add an error message to the `Err` field
+in the response. If no error then you can either send an empty response (`{}`)
+or an empty value for the `Err` field. Errors will only be logged.

docs/reference/commandline/plugin_ls.md

@@ -55,7 +55,7 @@ than one filter, then pass multiple flags (e.g., `--filter "foo=bar" --filter "b
 The currently supported filters are:
 
 * enabled (boolean - true or false, 0 or 1)
-* capability (string - currently `volumedriver`, `networkdriver`, `ipamdriver`, or `authz`)
+* capability (string - currently `volumedriver`, `networkdriver`, `ipamdriver`, `logdriver`, `metricscollector`, or `authz`)
 
 #### enabled
 
@@ -65,7 +65,7 @@ The `enabled` filter matches on plugins enabled or disabled.
 
 The `capability` filter matches on plugin capabilities. One plugin
 might have multiple capabilities. Currently `volumedriver`, `networkdriver`,
-`ipamdriver`, and `authz` are supported capabilities.
+`ipamdriver`, `logdriver`, `metricscollector`, and `authz` are supported capabilities.
 
 ```bash
 $ docker plugin install --disable tiborvass/no-remove