feat: collect process metrics (#12)

* feat: event loop utilization in telemetry

* release: v0.5.0

Author: Simone Sanfratello

README.md

@@ -44,6 +44,10 @@ telemetry.resetDurations()
 telemetry.resetCounters()
 ```
 
+#### Options
+
+- `configFile`: path to yml config file, see [Configuration file format](#configuration-file-format)
+
 #### Telemetry instance methods
 * export: Export the metrics in `prometheus` format
   ```
@@ -52,11 +56,14 @@ telemetry.resetCounters()
     counter_label_count_total{id="123"} 1 now
     counter_label_count_total{id="456"} 2 now
     ```
+* get(name): Get a single metric
+* getGaugeValue(name): Get the value of the gauge metric
+* getHistogramValue(name): Get the value (sum) of the histogram metric
 * resetAll(): Reset all metrics
 * resetCounters(): Reset count and labelCount metrics
 * resetDurations(): Reset durations metrics
 * resetGauges(): Reset gauges metrics
-* async export(): Export values  in Prometheus format 
+* async export(): Export values  in Prometheus format
 * increaseCount(category, amount = 1): Increase the count for a category
   * category: String - The given name of the category
 * increaseLabelCount(category, labels: Array, amount = 1): Increase the count for a key in a category
@@ -76,9 +83,10 @@ telemetry.resetCounters()
   * category: String - The given name of the category
   * promise: Promise - The function to be tracked
 
-#### Configuration
+#### Configuration file format
+
+Eg. of `metrics.yml`, to be passed to `configFile`
 
-Eg. of `metrics.yml`
 ```yaml
 ---
 component: bitswap-peer
@@ -102,6 +110,13 @@ metrics:
       description: BitSwap Event Loop Utilization
     bitswap-total-connections:
       description: BitSwap Total Connections
+
+  process:
+    elu: 
+      name: bitswap-elu
+      description: Bitswap Event Loop Utilization
+      interval: 500
+
 version: 0.1.0
 buildDate: "20220307.1423"
 ```

package.json

@@ -1,6 +1,6 @@
 {
   "name": "e-ipfs-core-lib",
-  "version": "0.4.0",
+  "version": "0.5.0",
   "description": "E-IPFS core library",
   "license": "(Apache-2.0 AND MIT)",
   "homepage": "https://github.com/elastic-ipfs/core-lib",

src/telemetry.js

@@ -1,7 +1,9 @@
+
+import { performance } from 'node:perf_hooks'
 import { readFileSync } from 'fs'
 import { load as ymlLoad } from 'js-yaml'
 import promClient from 'prom-client'
-const PERCENTILES = [0.001, 0.01, 0.1, 1, 2.5, 10, 25, 50, 75, 90, 97.5, 99, 99.9, 99.99, 99.999]
+const PERCENTILES = [0.1, 1, 10, 25, 50, 75, 90, 97.5, 99]
 
 class Telemetry {
   constructor ({ configFile, logger }) {
@@ -25,40 +27,46 @@ class Telemetry {
       // Setup
       this.component = component
       this.version = `${version}-build.${buildDate}`
+      this.metrics = new Map()
 
       // Create metrics
       for (const [category, metric] of Object.entries(metrics.count || {})) {
-      /* eslint-disable-next-line no-new */
-        new promClient.Counter({
+        this.metrics.set(category, new promClient.Counter({
           name: category.replaceAll('-', '_'),
           help: metric.description,
           registers: [this.countRegistry, this.allRegistry] // specify a non-default registry
-        })
+        }))
       }
       for (const [category, metric] of Object.entries(metrics.labelCount || {})) {
-        /* eslint-disable-next-line no-new */
-        new promClient.Counter({
+        this.metrics.set(category, new promClient.Counter({
           name: category.replaceAll('-', '_'),
           help: metric.description,
           labelNames: metric.labels,
           registers: [this.labelCountRegistry, this.allRegistry] // specify a non-default registry
-        })
+        }))
       }
       for (const [category, metric] of Object.entries(metrics.durations || {})) {
-        /* eslint-disable-next-line no-new */
-        new promClient.Histogram({
+        this.metrics.set(category, new promClient.Histogram({
           name: category.replaceAll('-', '_'),
           help: metric.description,
           buckets: PERCENTILES,
           registers: [this.durationsRegistry, this.allRegistry] // specify a non-default registry
-        })
+        }))
       }
       for (const [category, metric] of Object.entries(metrics.gauge || {})) {
-        /* eslint-disable-next-line no-new */
-        new promClient.Gauge({
+        this.metrics.set(category, new promClient.Gauge({
           name: category.replaceAll('-', '_'),
           help: metric.description,
           registers: [this.gaugeRegistry, this.allRegistry] // specify a non-default registry
+        }))
+      }
+
+      if (metrics.process?.elu) {
+        this.collectEventLoopUtilization({
+          name: metrics.process.elu.name,
+          description: metrics.process.elu.description,
+          interval: metrics.process.elu.interval,
+          registers: [this.allRegistry]
         })
       }
     } catch (err) {
@@ -67,6 +75,23 @@ class Telemetry {
     }
   }
 
+  collectEventLoopUtilization ({ name, description, interval, registers }) {
+    const metric = new promClient.Gauge({
+      name: name.replaceAll('-', '_'),
+      help: description,
+      registers
+    })
+    this.metrics.set(name, metric)
+
+    let elu1 = performance.eventLoopUtilization()
+    this.collectEluInterval = setInterval(() => {
+      const elu2 = performance.eventLoopUtilization()
+      const u = performance.eventLoopUtilization(elu2, elu1)
+      metric.set(u.utilization)
+      elu1 = elu2
+    }, interval).unref()
+  }
+
   resetAll () {
     this.allRegistry.resetMetrics()
   }
@@ -88,38 +113,54 @@ class Telemetry {
     return this.allRegistry.metrics()
   }
 
+  get (name) {
+    return this.metrics.get(name)
+  }
+
+  getGaugeValue (name) {
+    const value = this.get(name)
+    if (!value) { return }
+    return value.hashMap[''].value
+  }
+
+  getHistogramValue (name) {
+    const value = this.get(name)
+    if (!value) { return }
+    return value.hashMap && value.hashMap['']?.sum
+  }
+
   increaseCount (category, amount = 1) {
-    const metric = this.countRegistry.getSingleMetric(category.replaceAll('-', '_'))
+    const metric = this.metrics.get(category)
     if (!metric) throw new Error(`Metric ${category} not found`)
     return metric.inc(amount)
   }
 
   increaseLabelCount (category, labels = [], amount = 1) {
-    const metric = this.labelCountRegistry.getSingleMetric(category.replaceAll('-', '_'))
+    const metric = this.metrics.get(category)
     if (!metric) throw new Error(`Metric ${category} not found`)
     return metric.labels(...labels).inc(amount)
   }
 
   increaseGauge (category, amount = 1) {
-    const metric = this.gaugeRegistry.getSingleMetric(category.replaceAll('-', '_'))
+    const metric = this.metrics.get(category)
     if (!metric) throw new Error(`Metric ${category} not found`)
     return metric.inc(amount)
   }
 
   decreaseGauge (category, amount = 1) {
-    const metric = this.gaugeRegistry.getSingleMetric(category.replaceAll('-', '_'))
+    const metric = this.metrics.get(category)
     if (!metric) throw new Error(`Metric ${category} not found`)
     return metric.inc(-1 * amount)
   }
 
   setGauge (category, value) {
-    const metric = this.gaugeRegistry.getSingleMetric(category.replaceAll('-', '_'))
+    const metric = this.metrics.get(category)
     if (!metric) throw new Error(`Metric ${category} not found`)
     return metric.set(value)
   }
 
   async trackDuration (category, promise) {
-    const metric = this.durationsRegistry.getSingleMetric(category.replaceAll('-', '_'))
+    const metric = this.metrics.get(category)
     if (!metric) throw new Error(`Metric ${category} not found`)
 
     const end = metric.startTimer()

test/fixtures/process-metrics.yml

@@ -0,0 +1,15 @@
+---
+component: bitswap-peer
+metrics:
+  count:
+    s3-request-count:
+      description: AWS S3 requests
+
+  process:
+    elu: 
+      name: bitswap-elu
+      description: Bitswap Event Loop Utilization
+      interval: 100
+
+version: 0.2.0
+buildDate: "20220307.1423"