Merge pull request #4152 from specklesystems/iain/fix-prometheus-summary

fix(server): prometheus summary should be over rolling window

packages/server/modules/previews/observability/metrics.ts

@@ -1,3 +1,4 @@
+import { TIME } from '@speckle/shared'
 import Bull from 'bull'
 import { type Registry, Counter, Summary, Gauge } from 'prom-client'
 
@@ -107,8 +108,10 @@ export const initializeMetrics = (params: {
   )
   const previewJobsProcessedSummary = new Summary<'status'>({
     name: 'speckle_server_preview_jobs_processed_duration_seconds',
-    help: 'Duration of preview job processing, in seconds',
-    labelNames: ['status']
+    help: 'Duration of preview job processing, in seconds, as sampled over a period of 1 minute.',
+    labelNames: ['status'],
+    maxAgeSeconds: 1 * TIME.minute,
+    ageBuckets: 5
   })
 
   return { previewJobsProcessedSummary }

packages/server/observability/components/knex/knexMonitoring.ts

@@ -5,7 +5,7 @@ import { Logger } from 'pino'
 import { toNDecimalPlaces } from '@/modules/core/utils/formatting'
 import { omit } from 'lodash'
 import { getRequestContext } from '@/observability/components/express/requestContext'
-import { collectLongTrace } from '@speckle/shared'
+import { collectLongTrace, TIME } from '@speckle/shared'
 
 let metricQueryDuration: Summary<string>
 let metricQueryErrors: Counter<string>
@@ -134,7 +134,9 @@ export const initKnexPrometheusMetrics = async (params: {
       registers,
       labelNames: ['sqlMethod', 'sqlNumberBindings', 'region'],
       name: 'speckle_server_knex_query_duration',
-      help: 'Summary of the DB query durations in seconds'
+      help: 'Summary of the DB query durations in seconds, as computed over the last 1 minute.',
+      maxAgeSeconds: 1 * TIME.minute,
+      ageBuckets: 5
     })
 
     registers.forEach((r) => r.removeSingleMetric('speckle_server_knex_query_errors'))