Merge pull request #139 from davidnewhall/dn2_prometheus

Add prometheus exporter.

.metadata.sh

@@ -35,7 +35,7 @@ URL="${SOURCE_URL}"
 # This parameter is passed in as -X to go build. Used to override the Version variable in a package.
 # This makes a path like github.com/user/hello-world/helloworld.Version=1.3.3
 # Name the Version-containing library the same as the github repo, without dashes.
-VERSION_PATH="${IMPORT_PATH}/$(echo ${BINARY} | tr -d -- -).Version"
+VERSION_PATH="${IMPORT_PATH}/poller.Version"
 
 # Dynamic. Recommend not changing.
 VVERSION=$(git describe --abbrev=0 --tags $(git rev-list --tags --max-count=1))

examples/MANUAL.md

@@ -75,7 +75,8 @@ is provided so the application can be easily adapted to any environment.
 
     interval               default: 30s
         How often to poll the controller for updated client and device data.
-        The UniFi Controller only updates traffic stats about every 30 seconds.
+        The UniFi Controller only updates traffic stats about every 30-60 seconds.
+        Only works if "mode" (below) is "influx" - other modes do not use interval.
 
     debug                  default: false
         This turns on time stamps and line numbers in logs, outputs a few extra
@@ -89,10 +90,10 @@ is provided so the application can be easily adapted to any environment.
     mode                   default: "influx"
         * Value: influx
         This default mode runs this application as a daemon. It will poll
-        the controller at the configured interval. Providing an invalid value
-        will run in this default mode.
+        the controller at the configured interval and report measurements to
+        InfluxDB. Providing an invalid value will run in this default mode.
 
-        * Value: influxlambda  -  (the only other available option right now)
+        * Value: influxlambda
         Setting this value will invoke a run-once mode where the application
         immediately polls the controller and reports the metrics to InfluxDB.
         Then it exits. This mode is useful in an AWS Lambda or a crontab where
@@ -101,6 +102,16 @@ is provided so the application can be easily adapted to any environment.
         This mode can also be combined with a "test database" in InfluxDB to
         give yourself a "test config file" you may run ad-hoc to test changes.
 
+        * Value: prometheus
+        In this mode the application opens an http interface and exports the
+        measurements at /metrics for collection by prometheus. Enabling this
+        mode disables InfluxDB usage entirely.
+
+    http_listen            default: 0.0.0.0:9130
+        This option controls the IP and port the http listener uses when the
+        mode is set to prometheus. This setting has no effect when other modes
+        are in use. Metrics become available at the /metrics URI.
+
     influx_url             default: http://127.0.0.1:8086
         This is the URL where the Influx web server is available.
 

examples/up.conf.example

@@ -8,7 +8,8 @@
 sites = ["all"]
 
 # The UniFi Controller only updates traffic stats about every 30 seconds.
-# Setting this to something lower may lead to "zeros" in your data. You've been warned.
+# Setting this to something lower may lead to "zeros" in your data.
+# If you're getting zeros now, set this to "1m"
 interval = "30s"
 
 # Turns on line numbers, microsecond logging, and a per-device log.
@@ -24,13 +25,22 @@ quiet = false
 # an invalid mode will also result in "influx". In this default mode the application
 # runs as a daemon and polls the controller at the configured interval.
 #
-# There is only one other option at this time: "influxlambda"
+# There are two other options at this time: "influxlambda" and "prometheus"
 #
-# Lambda mode makes the application exit after collecting and reporting metrics
+# Mode "influxlambda" makes the application exit after collecting and reporting metrics
 # to InfluxDB one time. This mode requires an external process like an AWS Lambda
 # or a simple crontab to keep the timings accurate on UniFi Poller run intervals.
+#
+# Mode "prometheus" opens an HTTP server on port 9130 and exports the metrics at
+# /metrics for polling collection by a prometheus server. This disables influxdb.
+# IMPORTANT: The prometheus mode is still beta.
+# Please help us test and provide your feedback on the github repo!
 mode = "influx"
 
+# This controls on which ip and port /metrics is exported when mode is "prometheus".
+# This has no effect in other modes. Must contain a colon and port.
+http_listen = "0.0.0.0:9130"
+
 # InfluxDB does not require auth by default, so the user/password are probably unimportant.
 influx_url = "http://127.0.0.1:8086"
 influx_user = "unifi"

examples/up.json.example

@@ -4,6 +4,7 @@
  "debug": false,
  "quiet": false,
  "mode": "influx",
+ "http_listen": "0.0.0.0:9130",
  "influx_url": "http://127.0.0.1:8086",
  "influx_user": "unifi",
  "influx_pass": "unifi",

examples/up.xml.example

@@ -18,7 +18,8 @@
 
   <!--
   # The UniFi Controller only updates traffic stats about every 30 seconds.
-  # Setting this to something lower may lead to "zeros" in your data. You've been warned.
+  # Setting this to something lower may lead to "zeros" in your data.
+  # If you're getting zeros now, set this to "1m"
   -->
   <interval>30s</interval>
 
@@ -40,14 +41,25 @@
   # an invalid mode will also result in "influx". In this default mode the application
   # runs as a daemon and polls the controller at the configured interval.
   #
-  # There is only one other option at this time: "influxlambda"
+  # There are two other options at this time: "influxlambda" and "prometheus"
   #
-  # Lambda mode makes the application exit after collecting and reporting metrics
+  # Mode "influxlambda" makes the application exit after collecting and reporting metrics
   # to InfluxDB one time. This mode requires an external process like an AWS Lambda
   # or a simple crontab to keep the timings accurate on UniFi Poller run intervals.
+  #
+  # Mode "prometheus" opens an HTTP server on port 9130 and exports the metrics at
+  # /metrics for polling collection by a prometheus server. This disables influxdb.
+  # IMPORTANT: The prometheus mode is still beta.
+  # Please help us test and provide your feedback on the github repo!
   -->
   <mode>influx</mode>
 
+  <!--
+  # This controls on which ip and port /metrics is exported when mode is "prometheus".
+  # This has no effect in other modes. Must contain a colon and port.
+  -->
+  <http_listen>0.0.0.0:9130</http_listen>
+
   <!--
   # InfluxDB does not require auth by default, so the user/password are probably unimportant.
   -->

examples/up.yaml.example

@@ -9,7 +9,9 @@ sites:
   - all
 
 # The UniFi Controller only updates traffic stats about every 30 seconds.
-# Setting this to something lower may lead to "zeros" in your data. You've been warned.
+# Setting this to something lower may lead to "zeros" in your data.
+# If you're getting zeros now, set this to "1m"
+# Only has effect if "mode" (below) is "influx" - other modes do not use interval.
 interval: "30s"
 
 # Turns on line numbers, microsecond logging, and a per-device log.
@@ -25,13 +27,22 @@ quiet: false
 # an invalid mode will also result in "influx". In this default mode the application
 # runs as a daemon and polls the controller at the configured interval.
 #
-# There is only one other option at this time: "influxlambda"
+# There are two other options at this time: "influxlambda" and "prometheus"
 #
-# Lambda mode makes the application exit after collecting and reporting metrics
+# Mode "influxlambda" makes the application exit after collecting and reporting metrics
 # to InfluxDB one time. This mode requires an external process like an AWS Lambda
 # or a simple crontab to keep the timings accurate on UniFi Poller run intervals.
+#
+# Mode "prometheus" opens an HTTP server on port 9130 and exports the metrics at
+# /metrics for polling collection by a prometheus server. This disables influxdb.
+# IMPORTANT: The prometheus mode is still beta.
+# Please help us test and provide your feedback on the github repo!
 mode: "influx"
 
+# This controls on which ip and port /metrics is exported when mode is "prometheus".
+# This has no effect in other modes. Must contain a colon and port.
+http_listen: "0.0.0.0:9130"
+
 # InfluxDB does not require auth by default, so the user/password are probably unimportant.
 influx_url: "http://127.0.0.1:8086"
 influx_user: "unifi"

influxunifi/README.md

@@ -0,0 +1,4 @@
+# influx
+
+This package provides the methods to turn UniFi measurements into influx
+data-points with appropriate tags and fields.

unifipoller/influx_clients.go → influxunifi/clients.go

@@ -1,4 +1,4 @@
-package unifipoller
+package influxunifi
 
 import (
 	"time"

unifipoller/influx_ids.go → influxunifi/ids.go

@@ -1,4 +1,4 @@
-package unifipoller
+package influxunifi
 
 import (
 	influx "github.com/influxdata/influxdb1-client/v2"

influxunifi/metrics.go

@@ -0,0 +1,69 @@
+// Package influx provides the methods to turn UniFi measurements into influx
+// data-points with appropriate tags and fields.
+package influxunifi
+
+import (
+	"github.com/davidnewhall/unifi-poller/metrics"
+	client "github.com/influxdata/influxdb1-client/v2"
+)
+
+// Metrics contains all the data from the controller and an influx endpoint to send it to.
+type Metrics struct {
+	*metrics.Metrics
+	client.BatchPoints
+}
+
+// ProcessPoints batches all device and client data into influxdb data points.
+// Call this after you've collected all the data you care about.
+// This function is sorta weird and returns a slice of errors. The reasoning is
+// that some points may process while others fail, so we attempt to process them
+// all. This is (usually) run in a loop, so we can't really exit on error,
+// we just log the errors and tally them on a counter. In reality, this never
+// returns any errors because we control the data going in; cool right? But we
+// still check&log it in case the data going is skewed up and causes errors!
+func (m *Metrics) ProcessPoints() []error {
+	errs := []error{}
+	processPoints := func(m *Metrics, p []*client.Point, err error) {
+		switch {
+		case err != nil:
+			errs = append(errs, err)
+		case p == nil:
+		default:
+			m.AddPoints(p)
+		}
+	}
+
+	for _, asset := range m.Sites {
+		pts, err := SitePoints(asset, m.TS)
+		processPoints(m, pts, err)
+	}
+	for _, asset := range m.Clients {
+		pts, err := ClientPoints(asset, m.TS)
+		processPoints(m, pts, err)
+	}
+	for _, asset := range m.IDSList {
+		pts, err := IDSPoints(asset) // no m.TS.
+		processPoints(m, pts, err)
+	}
+
+	if m.Devices == nil {
+		return errs
+	}
+	for _, asset := range m.Devices.UAPs {
+		pts, err := UAPPoints(asset, m.TS)
+		processPoints(m, pts, err)
+	}
+	for _, asset := range m.Devices.USGs {
+		pts, err := USGPoints(asset, m.TS)
+		processPoints(m, pts, err)
+	}
+	for _, asset := range m.Devices.USWs {
+		pts, err := USWPoints(asset, m.TS)
+		processPoints(m, pts, err)
+	}
+	for _, asset := range m.Devices.UDMs {
+		pts, err := UDMPoints(asset, m.TS)
+		processPoints(m, pts, err)
+	}
+	return errs
+}

unifipoller/influx_site.go → influxunifi/site.go

@@ -1,4 +1,4 @@
-package unifipoller
+package influxunifi
 
 import (
 	"strings"

unifipoller/influx_uap.go → influxunifi/uap.go

@@ -1,4 +1,4 @@
-package unifipoller
+package influxunifi
 
 import (
 	"time"

unifipoller/influx_udm.go → influxunifi/udm.go

@@ -1,4 +1,4 @@
-package unifipoller
+package influxunifi
 
 import (
 	"time"