chore(release): pull hotfix-release/v1.67.0 into main (#3410)

CHANGELOG.md

@@ -2,6 +2,13 @@
 
 All notable changes to this project will be documented in this file. See [standard-version](https://github.com/conventional-changelog/standard-version) for commit guidelines.
 
+## [1.67.0](https://github.com/rudderlabs/rudder-transformer/compare/v1.66.1...v1.67.0) (2024-05-23)
+
+
+### Features
+
+* sre 456 ut move high cardinality histogram metrics to summaries cp ([#3409](https://github.com/rudderlabs/rudder-transformer/issues/3409)) ([be20dc2](https://github.com/rudderlabs/rudder-transformer/commit/be20dc26ade2fa0212dc91126cf42087a84a07c9))
+
 ### [1.66.1](https://github.com/rudderlabs/rudder-transformer/compare/v1.66.0...v1.66.1) (2024-05-20)
 
 

package.json

@@ -1,6 +1,6 @@
 {
   "name": "rudder-transformer",
-  "version": "1.66.1",
+  "version": "1.67.0",
   "description": "",
   "homepage": "https://github.com/rudderlabs/rudder-transformer#readme",
   "bugs": {

src/legacy/router.js

@@ -649,6 +649,9 @@ if (startDestTransformer) {
       stats.timing('user_transform_request_latency', startTime, {
         processSessions,
       });
+      stats.timingSummary('user_transform_request_latency_summary', startTime, {
+        processSessions,
+      });
       stats.increment('user_transform_requests', { processSessions });
       stats.histogram('user_transform_output_events', transformedEvents.length, {
         processSessions,

src/services/userTransform.ts

@@ -173,7 +173,17 @@ export class UserTransformService {
             ...getTransformationMetadata(eventsToProcess[0]?.metadata),
           });
 
-          stats.histogram('user_transform_batch_size', requestSize, {
+          stats.timing('user_transform_batch_size', requestSize, {
+            ...metaTags,
+            ...getTransformationMetadata(eventsToProcess[0]?.metadata),
+          });
+
+          stats.timingSummary('user_transform_request_latency_summary', userFuncStartTime, {
+            ...metaTags,
+            ...getTransformationMetadata(eventsToProcess[0]?.metadata),
+          });
+
+          stats.timingSummary('user_transform_batch_size_summary', requestSize, {
             ...metaTags,
             ...getTransformationMetadata(eventsToProcess[0]?.metadata),
           });

src/util/customTransformer-v1.js

@@ -93,6 +93,7 @@ async function userTransformHandlerV1(
     };
     stats.counter('user_transform_function_input_events', events.length, tags);
     stats.timing('user_transform_function_latency', invokeTime, tags);
+    stats.timingSummary('user_transform_function_latency_summary', invokeTime, tags);
   }
 
   return { transformedEvents, logs };

src/util/customTransformer.js

@@ -260,6 +260,7 @@ async function runUserTransform(
 
     stats.counter('user_transform_function_input_events', events.length, tags);
     stats.timing('user_transform_function_latency', invokeTime, tags);
+    stats.timingSummary('user_transform_function_latency_summary', invokeTime, tags);
   }
 
   return {

src/util/customTransforrmationsStore-v1.js

@@ -31,6 +31,7 @@ async function getTransformationCodeV1(versionId) {
     responseStatusHandler(response.status, 'Transformation', versionId, url);
     stats.increment('get_transformation_code', { success: 'true', ...tags });
     stats.timing('get_transformation_code_time', startTime, tags);
+    stats.timingSummary('get_transformation_code_time_summary', startTime, tags);
     const myJson = await response.json();
     transformationCache[versionId] = myJson;
     return myJson;
@@ -56,6 +57,7 @@ async function getLibraryCodeV1(versionId) {
     responseStatusHandler(response.status, 'Transformation Library', versionId, url);
     stats.increment('get_libraries_code', { success: 'true', ...tags });
     stats.timing('get_libraries_code_time', startTime, tags);
+    stats.timingSummary('get_libraries_code_time_summary', startTime, tags);
     const myJson = await response.json();
     libraryCache[versionId] = myJson;
     return myJson;
@@ -83,6 +85,7 @@ async function getRudderLibByImportName(importName) {
     responseStatusHandler(response.status, 'Rudder Library', importName, url);
     stats.increment('get_libraries_code', { success: 'true', ...tags });
     stats.timing('get_libraries_code_time', startTime, tags);
+    stats.timingSummary('get_libraries_code_time_summary', startTime, tags);
     const myJson = await response.json();
     rudderLibraryCache[importName] = myJson;
     return myJson;

src/util/customTransforrmationsStore.js

@@ -24,6 +24,7 @@ async function getTransformationCode(versionId) {
     responseStatusHandler(response.status, 'Transformation', versionId, url);
     stats.increment('get_transformation_code', { versionId, success: 'true' });
     stats.timing('get_transformation_code_time', startTime, { versionId });
+    stats.timingSummary('get_transformation_code_time_summary', startTime, { versionId });
     const myJson = await response.json();
     myCache.set(versionId, myJson);
     return myJson;

src/util/openfaas/index.js

@@ -305,6 +305,7 @@ const executeFaasFunction = async (
 
     stats.counter('user_transform_function_input_events', events.length, tags);
     stats.timing('user_transform_function_latency', startTime, tags);
+    stats.timingSummary('user_transform_function_latency_summary', startTime, tags);
   }
 };
 

src/util/prometheus.js

@@ -11,7 +11,7 @@ function appendPrefix(name) {
 }
 
 class Prometheus {
-  constructor() {
+  constructor(enableSummaryMetrics = true) {
     this.prometheusRegistry = new prometheusClient.Registry();
     this.prometheusRegistry.setDefaultLabels(defaultLabels);
     prometheusClient.collectDefaultMetrics({
@@ -21,7 +21,7 @@ class Prometheus {
     prometheusClient.AggregatorRegistry.setRegistries(this.prometheusRegistry);
     this.aggregatorRegistry = new prometheusClient.AggregatorRegistry();
 
-    this.createMetrics();
+    this.createMetrics(enableSummaryMetrics);
   }
 
   async metricsController(ctx) {
@@ -56,11 +56,22 @@ class Prometheus {
     return gauge;
   }
 
-  newSummaryStat(name, help, labelNames) {
+  newSummaryStat(
+    name,
+    help,
+    labelNames,
+    percentiles = [0.5, 0.9, 0.99],
+    maxAgeSeconds = 300,
+    ageBuckets = 5,
+  ) {
+    // we enable a 5 minute sliding window and calculate the 50th, 90th, and 99th percentiles by default
     const summary = new prometheusClient.Summary({
       name,
       help,
       labelNames,
+      percentiles,
+      maxAgeSeconds,
+      ageBuckets,
     });
     this.prometheusRegistry.registerMetric(summary);
     return summary;
@@ -117,6 +128,21 @@ class Prometheus {
     }
   }
 
+  timingSummary(name, start, tags = {}) {
+    try {
+      let metric = this.prometheusRegistry.getSingleMetric(appendPrefix(name));
+      if (!metric) {
+        logger.warn(
+          `Prometheus: summary metric ${name} not found in the registry. Creating a new one`,
+        );
+        metric = this.newSummaryStat(name, name, Object.keys(tags));
+      }
+      metric.observe(tags, (new Date() - start) / 1000);
+    } catch (e) {
+      logger.error(`Prometheus: Summary metric ${name} failed with error ${e}`);
+    }
+  }
+
   histogram(name, value, tags = {}) {
     try {
       let metric = this.prometheusRegistry.getSingleMetric(appendPrefix(name));
@@ -166,7 +192,7 @@ class Prometheus {
     }
   }
 
-  createMetrics() {
+  createMetrics(enableSummaryMetrics) {
     const metrics = [
       // Counters
       {
@@ -698,6 +724,18 @@ class Prometheus {
           'k8_namespace',
         ],
       },
+      {
+        name: 'user_transform_request_latency_summary',
+        help: 'user_transform_request_latency_summary',
+        type: 'summary',
+        labelNames: [
+          'workspaceId',
+          'transformationId',
+          'sourceType',
+          'destinationType',
+          'k8_namespace',
+        ],
+      },
       {
         name: 'user_transform_batch_size',
         help: 'user_transform_batch_size',
@@ -714,6 +752,18 @@ class Prometheus {
           524288000,
         ], // 1KB, 100KB, 0.5MB, 1MB, 10MB, 20MB, 50MB, 100MB, 200MB, 500MB
       },
+      {
+        name: 'user_transform_batch_size_summary',
+        help: 'user_transform_batch_size_summary',
+        type: 'summary',
+        labelNames: [
+          'workspaceId',
+          'transformationId',
+          'sourceType',
+          'destinationType',
+          'k8_namespace',
+        ],
+      },
       {
         name: 'source_transform_request_latency',
         help: 'source_transform_request_latency',
@@ -770,12 +820,24 @@ class Prometheus {
         type: 'histogram',
         labelNames: ['versionId', 'version'],
       },
+      {
+        name: 'get_transformation_code_time_summary',
+        help: 'get_transformation_code_time_summary',
+        type: 'summary',
+        labelNames: ['versionId', 'version'],
+      },
       {
         name: 'get_libraries_code_time',
         help: 'get_libraries_code_time',
         type: 'histogram',
         labelNames: ['libraryVersionId', 'versionId', 'type', 'version'],
       },
+      {
+        name: 'get_libraries_code_time_summary',
+        help: 'get_libraries_code_time_summary',
+        type: 'summary',
+        labelNames: ['libraryVersionId', 'versionId', 'type', 'version'],
+      },
       {
         name: 'isolate_cpu_time',
         help: 'isolate_cpu_time',
@@ -1027,6 +1089,22 @@ class Prometheus {
           'workspaceId',
         ],
       },
+      {
+        name: 'user_transform_function_latency_summary',
+        help: 'user_transform_function_latency_summary',
+        type: 'summary',
+        labelNames: [
+          'identifier',
+          'testMode',
+          'sourceType',
+          'destinationType',
+          'k8_namespace',
+          'errored',
+          'statusCode',
+          'transformationId',
+          'workspaceId',
+        ],
+      },
     ];
 
     metrics.forEach((metric) => {
@@ -1042,6 +1120,17 @@ class Prometheus {
             metric.labelNames,
             metric.buckets,
           );
+        } else if (metric.type === 'summary') {
+          if (enableSummaryMetrics) {
+            this.newSummaryStat(
+              appendPrefix(metric.name),
+              metric.help,
+              metric.labelNames,
+              metric.percentiles,
+              metric.maxAge,
+              metric.ageBuckets,
+            );
+          }
         } else {
           logger.error(
             `Prometheus: Metric creation failed. Name: ${metric.name}. Invalid type: ${metric.type}`,

src/util/stats.js

@@ -4,6 +4,8 @@ const logger = require('../logger');
 
 const enableStats = process.env.ENABLE_STATS !== 'false';
 const statsClientType = process.env.STATS_CLIENT || 'statsd';
+// summary metrics are enabled by default. To disable set ENABLE_SUMMARY_METRICS='false'.
+const enableSummaryMetrics = process.env.ENABLE_SUMMARY_METRICS !== 'false';
 
 let statsClient;
 function init() {
@@ -19,7 +21,7 @@ function init() {
 
     case 'prometheus':
       logger.info('setting up prometheus client');
-      statsClient = new prometheus.Prometheus();
+      statsClient = new prometheus.Prometheus(enableSummaryMetrics);
       break;
 
     default:
@@ -38,6 +40,15 @@ const timing = (name, start, tags = {}) => {
   statsClient.timing(name, start, tags);
 };
 
+// timingSummary is used to record observations for a summary metric
+const timingSummary = (name, start, tags = {}) => {
+  if (!enableStats || !statsClient || !enableSummaryMetrics) {
+    return;
+  }
+
+  statsClient.timingSummary(name, start, tags);
+};
+
 const increment = (name, tags = {}) => {
   if (!enableStats || !statsClient) {
     return;
@@ -88,4 +99,13 @@ async function metricsController(ctx) {
 
 init();
 
-module.exports = { init, timing, increment, counter, gauge, histogram, metricsController };
+module.exports = {
+  init,
+  timing,
+  timingSummary,
+  increment,
+  counter,
+  gauge,
+  histogram,
+  metricsController,
+};

src/util/statsd.js

@@ -21,6 +21,11 @@ class Statsd {
     this.statsdClient.timing(name, start, tags);
   }
 
+  // timingSummary is just a wrapper around timing for statsd.For prometheus, we will have to implement a different function.
+  timingSummary(name, start, tags = {}) {
+    this.statsdClient.timing(name, start, tags);
+  }
+
   increment(name, tags = {}) {
     this.statsdClient.increment(name, 1, tags);
   }