documentation updates for upcoming release

docs/content/feature/metrics.md

@@ -11,6 +11,19 @@ to avoid computing large amounts of metrics. The metrics can be send to
 options in the [fabio.properties](https://github.com/eBay/fabio/blob/master/fabio.properties)
 file.
 
+### Configuring Prometheus Metrics
+
+To configure prometheus metrics, you need two do the following:
+
+1) You must specify that prometheus is the [metrics.target](/ref/metrics.target/)
+2) You must configure a listener with [proxy.addr](/ref/proxy.addr/) with `proto=prometheus`
+3) (optional) override the 
+[metrics.prometheus.path](/ref/metrics.prometheus.path/),
+[metrics.prometheus.subsystem](/ref/metrics.prometheus.subsystem/),
+and [metrics.prometheus.buckets](/ref/metrics.prometheus.buckets/). 
+
+### Metrics info (for non-tagged backends, such as circonus and statsd_raw)
+
 Fabio reports the following metrics:
 
 Name                        | Type     | Description
@@ -60,3 +73,4 @@ A gauge provides a current value.
 with the `metrics.names` template defined in
 [fabio.properties](https://github.com/fabiolb/fabio/blob/master/fabio.properties)
 
+

docs/content/ref/metrics.dogstatsd.addr.md

@@ -0,0 +1,12 @@
+---
+title: "metrics.dogstatsd.addr"
+---
+
+`metrics.dogstatsd.addr` configures the host:port of the dogstatsd
+server. 
+
+This is required when [metrics.target](/ref/metrics.target/) is set to `dogstatsd`.
+
+The default is
+
+	metrics.statsd.addr =

docs/content/ref/metrics.names.md

@@ -2,7 +2,9 @@
 title: "metrics.names"
 ---
 
-`metrics.names` configures the template for the route metric names.
+`metrics.names` configures the template for the route metric names
+on backends that don't support tags.  This is used in circonus,
+graphite and statsd_raw.  dogstatsd and prometheus ignore this.
 The value is expanded by the [text/template](https://golang.org/pkg/text/template) package and provides
 the following variables:
 

docs/content/ref/metrics.prometheus.buckets.md

@@ -0,0 +1,9 @@
+---
+title: "metrics.prometheus.buckets"
+---
+
+`metrics.prometheus.buckets` configures the time buckets for use with histograms, measured in seconds.
+for instance, .005 is equivalent to 5ms.  There is an implied "infinity" bucket tacked on at the end.
+
+The default is
+`metrics.prometheus.buckets = .005,.01,.025,.05,.1,.25,.5,1,2.5,5,10`

docs/content/ref/metrics.prometheus.path.md

@@ -0,0 +1,8 @@
+---
+title: "metrics.prometheus.path"
+---
+
+`metrics.prometheus.path` configures the path to serve up metrics on any configured
+[proxy.addr](/ref/proxy.addr/) where `proto=prometheus`.
+
+Defaults to `/metrics/`

docs/content/ref/metrics.prometheus.subsystem.md

@@ -0,0 +1,9 @@
+---
+title: "metrics.prometheus.subsystem"
+---
+
+`metrics.prometheus.subsystem` configures the subsystem name when reporting
+metrics.  This is basically appended to the prefix for metric names.
+
+See https://prometheus.io/docs/practices/instrumentation/#subsystems
+for more information.

docs/content/ref/metrics.statsd.addr.md

@@ -5,7 +5,7 @@ title: "metrics.statsd.addr"
 `metrics.statsd.addr` configures the host:port of the StatsD
 server. 
 
-This is required when [metrics.target](/ref/metrics.target/) is set to `statsd`.
+This is required when [metrics.target](/ref/metrics.target/) is set to `statsd_raw`.
 
 The default is
 

docs/content/ref/metrics.target.md

@@ -9,9 +9,18 @@ Possible values are:
 * `<empty>`:  do not report metrics
 * `stdout`:   report metrics to stdout
 * `graphite`: report metrics to Graphite on [metrics.graphite.addr](/ref/metrics.graphite.addr/)
-* `statsd`: report metrics to StatsD on [metrics.statsd.addr](/ref/metrics.statsd.addr/)
-* `circonus`: report metrics to Circonus (http://circonus.com/)
+* `statsd`: legacy statsd support, used in v1.5.5 and lower - removed in v1.6
+* `statsd_raw`: report metrics to StatsD on [metrics.statsd.addr](/ref/metrics.statsd.addr/) - this was 
+  intentionally renamed because anyone upgrading to 1.6 will need to revisit their configuration anyway due to 
+  rewrite of this backend.  It was quite broken before, the counters never reset, it did not follow the spec so the info was 
+  likely wrong or people using this were doing some workarounds they'll need to remove anyway.
+* `circonus`: report metrics to Circonus (https://circonus.com/)
+* `prometheus`: use prometheus metrics. (https://prometheus.io)  Must be used in conjuction with a prometheus 
+  listener in [proxy.addr](/ref/proxy.addr/])
+* `dogstatsd`: use with datadog dogstatsd (https://www.datadoghq.com/)
 
 The default is
 
 	metrics.target =
+
+Multiple metrics targets can be defined separated by comma.

docs/content/ref/proxy.addr.md

@@ -24,6 +24,8 @@ The supported protocols are:
 * `tcp+sni` for an SNI aware TCP proxy
 * `tcp-dynamic` for a consul driven TCP proxy
 * `https+tcp+sni` for an SNI aware TCP proxy with https fallthrough
+* `prometheus` for a prometheus metrics endpoint.  Used in conjunction with [metrics.target](/ref/metrics.target/)
+  =prometheus
 
 If no `proto` option is specified then the protocol
 is either `http` or `https` depending on whether a

docs/content/ref/proxy.idleconntimeout.md

@@ -0,0 +1,6 @@
+`proxy.idleconntimeout` configures idle connection timeout, which influences
+when to close keep-alive connections.
+
+The default is
+
+    proxy.idleconntimeout     = 15s

fabio.properties

@@ -375,6 +375,14 @@
 # proxy.keepalivetimeout     = 0s
 
 
+# proxy.idleconntimeout configures the idle connection timeout, when
+# to close (keep-alive) connections
+#
+# The default is
+#
+# proxy.idleconntimeout = 15s
+
+
 # proxy.dialtimeout configures the connection timeout for
 # outgoing connections.
 #