support and autoscaling additions

developmentseed · Aug 27, 2024 · cde90f5 · cde90f5
1 parent a092153
commit cde90f5
Show file tree

Hide file tree

Showing 6 changed files with 35 additions and 28 deletions.
diff --git a/.gitignore b/.gitignore
@@ -1,3 +1,4 @@
 .idea/
 helm-chart/config.yaml
-config_ingress.yaml
+config_ingress.yaml
+helm-chart/eoapi/charts/*.tgz
diff --git a/helm-chart/eoapi-support/values.yaml b/helm-chart/eoapi-support/values.yaml
@@ -20,23 +20,23 @@ prometheus-adapter:
       name:
         matches: ""
         as: "nginx_ingress_controller_requests_rate_vector_eoapi"
-      metricsQuery: round(sum(rate(<<.Series>>{service=~"vector.*",<<.LabelMatchers>>}[5m])) by (<<.GroupBy>>), 0.001)
+      metricsQuery: round(sum(rate(<<.Series>>{service="vector",path=~"/vector.*",<<.LabelMatchers>>}[5m])) by (<<.GroupBy>>), 0.001)
     - seriesQuery: '{__name__=~"^nginx_ingress_controller_requests$",namespace!=""}'
       seriesFilters: []
       resources:
         template: <<.Resource>>
       name:
         matches: ""
         as: "nginx_ingress_controller_requests_rate_raster_eoapi"
-      metricsQuery: round(sum(rate(<<.Series>>{service=~"raster.*",<<.LabelMatchers>>}[5m])) by (<<.GroupBy>>), 0.001)
+      metricsQuery: round(sum(rate(<<.Series>>{service="raster",path=~"/raster.*",<<.LabelMatchers>>}[5m])) by (<<.GroupBy>>), 0.001)
     - seriesQuery: '{__name__=~"^nginx_ingress_controller_requests$",namespace!=""}'
       seriesFilters: []
       resources:
         template: <<.Resource>>
       name:
         matches: ""
         as: "nginx_ingress_controller_requests_rate_stac_eoapi"
-      metricsQuery: round(sum(rate(<<.Series>>{service=~"stac.*",<<.LabelMatchers>>}[5m])) by (<<.GroupBy>>), 0.001)
+      metricsQuery: round(sum(rate(<<.Series>>{service="stac",path=~"/stac.*",<<.LabelMatchers>>}[5m])) by (<<.GroupBy>>), 0.001)
 
 prometheus:
   # alertmanager is an optional prometheus chart dependency that we opt-out from

diff --git a/helm-chart/eoapi/Chart.yaml b/helm-chart/eoapi/Chart.yaml
@@ -32,4 +32,4 @@ dependencies:
   - name: eoapi-support
     version: 0.1.4
     repository: https://devseed.com/eoapi-k8s/
-    condition: autoscalingCondition
+    condition: autoscaling.enabled
diff --git a/helm-chart/eoapi/templates/_helpers.tpl b/helm-chart/eoapi/templates/_helpers.tpl
@@ -218,12 +218,3 @@ that you cannot have db.enabled and (postgrescluster.enabled or pgstacBootstrap.
 {{- end -}}
 
 {{- end -}}
-
-
-{{- define "check_autoscaling_enabled" -}}
-{{- $enabled := false -}}
-{{- if or .Values.raster.autoscaling.enabled .Values.vector.autoscaling.enabled .Values.stac.autoscaling.enabled -}}
-  {{- $enabled = true -}}
-{{- end -}}
-{{- $enabled -}}
-{{- end -}}
diff --git a/helm-chart/eoapi/templates/services/ingress-nginx.yaml b/helm-chart/eoapi/templates/services/ingress-nginx.yaml
@@ -15,6 +15,8 @@ metadata:
     nginx.ingress.kubernetes.io/use-regex: "true"
     nginx.ingress.kubernetes.io/rewrite-target: /$2
     nginx.ingress.kubernetes.io/enable-cors: "true"
+    # enable-access-log is required for nginx to dump metrics about path rewrites for prometheus to scrape
+    nginx.ingress.kubernetes.io/enable-access-log: "true"
     {{- if (and (.Values.ingress.tls.enabled) (.Values.ingress.tls.certManager)) }}
     cert-manager.io/issuer: {{ .Values.ingress.tls.certManagerIssuer }}
     {{- end }}

diff --git a/helm-chart/eoapi/values.yaml b/helm-chart/eoapi/values.yaml
@@ -170,8 +170,12 @@ pgstacBootstrap:
 ######################
 # API SERVICES
 ######################
-# check if ANY of the services below enabled autoscaling and conditionally install the eoapi-support helm chart
-autoscalingCondition: {{ include "check_autoscaling_enabled" . }}
+# a flag that is only used in the eoapi chart dependencies condition block
+# to install the eoapi-support chart that includes the additional observability
+# since we cannot use any syntax that such as
+# (or .Values.raster.autoscaling.enabled .Values.vector.autoscaling.enabled .Values.stac.autoscaling.enabled)
+autoscaling:
+  enabled: false
 
 raster:
   enabled: true
@@ -180,16 +184,20 @@ raster:
     minReplicas: 1
     maxReplicas: 10
     # `type`: "cpu" || "requestRate" || "both"
-    type: "requestRate"
+    type: "both"
     behaviour:
       scaleDown:
         stabilizationWindowSeconds: 60
       scaleUp:
         stabilizationWindowSeconds: 0
     targets:
       # matches `type` value above unless `type: "both"` is selected
-      cpu: 85
-      requestRate: 10000
+      cpu: 75
+      # 'm' units here represents generic milli (one-thousandth) unit instead of 'decimal'
+      # https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale-walkthrough/#quantities
+      # so when the average unit among these pods is <requestRate>/1000 then scale
+      # you can watch the actual/target in real time using `kubectl get hpa/<name>`
+      requestRate: 10000m
   image:
     name: ghcr.io/stac-utils/titiler-pgstac
     tag: uvicorn-1.2.0
@@ -236,16 +244,20 @@ stac:
     minReplicas: 1
     maxReplicas: 10
     # `type`: "cpu" || "requestRate" || "both"
-    type: "requestRate"
+    type: "both"
     behaviour:
       scaleDown:
         stabilizationWindowSeconds: 60
       scaleUp:
         stabilizationWindowSeconds: 0
     targets:
       # matches `type` value above unless `type: "both"` is selected
-      cpu: 85
-      requestRate: 15000
+      cpu: 75
+      # 'm' units here represents generic milli (one-thousandth) unit instead of 'decimal'
+      # https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale-walkthrough/#quantities
+      # so when the average unit among these pods is <requestRate>/1000 then scale
+      # you can watch the actual/target in real time using `kubectl get hpa/<name>`
+      requestRate: 60000m
   image:
     name: ghcr.io/stac-utils/stac-fastapi-pgstac
     tag: 2.4.9
@@ -279,16 +291,20 @@ vector:
     minReplicas: 1
     maxReplicas: 10
     # `type`: "cpu" || "requestRate" || "both"
-    type: "requestRate"
+    type: "both"
     behaviour:
       scaleDown:
         stabilizationWindowSeconds: 60
       scaleUp:
         stabilizationWindowSeconds: 0
     targets:
       # matches `type` value above unless `type: "both"` is selected
-      cpu: 85
-      requestRate: 15000
+      cpu: 75
+      # 'm' units here represents generic milli (one-thousandth) unit instead of 'decimal'
+      # https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale-walkthrough/#quantities
+      # so when the average unit among these pods is <requestRate>/1000 then scale
+      # you can watch the actual/target in real time using `kubectl get hpa/<name>`
+      requestRate: 60000m
   image:
     name: ghcr.io/developmentseed/tipg
     tag: uvicorn-0.6.1
@@ -322,6 +338,3 @@ vector:
 
 docServer:
   enabled: true
-
-
-