Issue #3968: Allow hostNetwork=true for seldon operator (#3971)

* Issue #3968: Allow hostNetwork=true for seldon operator * Issue #3968: Allow hostNetwork=true for seldon operator update helm template generators * Issue #3968: Allow hostNetwork=true for seldon operator remove duplicated code * Issue #3968: Allow hostNetwork=true for seldon operator Remove quotes from hostNetwork when generating helm chart * Issue #3968: Allow hostNetwork=true for seldon operator Remove quotes from hostNetwork when generating helm chart Co-authored-by: Mariano Billinghurst <mariano.billinghurst@pmi.com>
SeldonIO · Mar 17, 2022 · c515a00 · c515a00
1 parent 845e809
commit c515a00
Show file tree

Hide file tree

Showing 6 changed files with 47 additions and 16 deletions.
diff --git a/helm-charts/seldon-core-operator/README.md b/helm-charts/seldon-core-operator/README.md
@@ -117,3 +117,5 @@ helm install seldon-core-operator seldonio/seldon-core-operator --namespace seld
 | storageInitializer.memoryRequest | string | `"100Mi"` |  |
 | usageMetrics.enabled | bool | `false` |  |
 | webhook.port | int | `4443` |  |
+| metrics.port | int | `8080` |  |
+| hostNetwork | bool | `false` |  |
diff --git a/helm-charts/seldon-core-operator/templates/deployment_seldon-controller-manager.yaml b/helm-charts/seldon-core-operator/templates/deployment_seldon-controller-manager.yaml
@@ -24,6 +24,7 @@ spec:
 {{- with .Values.manager.annotations }}
 {{- toYaml . | nindent 8}}
 {{- end }}
+        prometheus.io/port: '{{ .Values.metrics.port }}'
         prometheus.io/scrape: 'true'
         sidecar.istio.io/inject: 'false'
       labels:
@@ -140,7 +141,7 @@ spec:
         - containerPort: {{ .Values.webhook.port }}
           name: webhook-server
           protocol: TCP
-        - containerPort: 8080
+        - containerPort: {{ .Values.metrics.port }}
           name: metrics
           protocol: TCP
         resources:
@@ -156,6 +157,7 @@ spec:
           name: cert
           readOnly: true
 {{- end }}
+      hostNetwork: {{ .Values.hostNetwork }}
       securityContext:
         runAsUser: {{ .Values.managerUserID }}
       serviceAccountName: '{{ .Values.serviceAccount.name }}'

diff --git a/helm-charts/seldon-core-operator/values.yaml b/helm-charts/seldon-core-operator/values.yaml
@@ -2,7 +2,7 @@
 # Below are the default values when installing Seldon Core
 
 # Defaults to .Release.Namespace
-namespaceOverride: "" 
+namespaceOverride: ""
 
 # ## Ingress Options
 # You are able to choose between Istio and Ambassador
@@ -69,6 +69,7 @@ executor:
     workQueueSize: 10000
     writeTimeoutMs: 2000
 
+
 # ## Seldon Core Controller Manager Options
 image:
   pullPolicy: IfNotPresent
@@ -101,13 +102,18 @@ storageInitializer:
   memoryRequest: 100Mi
 usageMetrics:
   enabled: false
+# In scenarios like EKS with non-standard CNI plugin like calico, the control plane cannot reach the webhook
+# hence it is needed to set hostNetwork: true
+hostNetwork: false
 webhook:
-  port: 4443
+  port: 4443 # If 'hostNetwork: true' you might need to change this port if it is already used by the node
+metrics: # these are the metrics exposed by the controller pod
+  port: 8080 # If 'hostNetwork: true' you might need to change this port if it is already used by the node
 
 # ## Predictive Unit Values
 predictiveUnit:
   httpPort: 9000
-  grpcPort: 9500  
+  grpcPort: 9500
   metricsPortName: metrics
   # If you would like to add extra environment variables to the init container to make available
   #   secrets such as cloud credentials, you can provide a default secret name that will be loaded
@@ -135,7 +141,7 @@ predictor_servers:
       seldon:
         defaultImageVersion: "1.14.0-dev"
         image: seldonio/tfserving-proxy
-      tensorflow: 
+      tensorflow:
         defaultImageVersion: 2.1.0
         image:  tensorflow/serving
   XGBOOST_SERVER:

diff --git a/operator/config/default/manager_prometheus_metrics_patch.yaml b/operator/config/default/manager_prometheus_metrics_patch.yaml
@@ -9,6 +9,7 @@ spec:
     metadata:
       annotations:
         prometheus.io/scrape: 'true'
+        prometheus.io/port: '8080'
     spec:
       containers:
       # Expose the prometheus metrics on default port

diff --git a/operator/config/manager/manager.yaml b/operator/config/manager/manager.yaml
@@ -24,6 +24,7 @@ spec:
       labels:
         control-plane: seldon-controller-manager
     spec:
+      hostNetwork: false
       serviceAccountName: manager
       securityContext:
         runAsUser: 8888
@@ -38,30 +39,30 @@ spec:
         - --leader-election-id=$(MANAGER_LEADER_ELECTION_ID)
         env:
         - name: MANAGER_LEADER_ELECTION_ID
-          value: "a33bd623.machinelearning.seldon.io"        
+          value: "a33bd623.machinelearning.seldon.io"
         - name: MANAGER_LOG_LEVEL
           value: "INFO"
         - name: WATCH_NAMESPACE
           value: ""
         - name: RELATED_IMAGE_EXECUTOR
           value: ""
         - name: RELATED_IMAGE_STORAGE_INITIALIZER
-          value: ""          
+          value: ""
         - name: RELATED_IMAGE_SKLEARNSERVER
-          value: ""          
+          value: ""
         - name: RELATED_IMAGE_XGBOOSTSERVER
-          value: ""          
+          value: ""
         - name: RELATED_IMAGE_MLFLOWSERVER
-          value: ""          
+          value: ""
         - name: RELATED_IMAGE_TFPROXY
-          value: ""          
+          value: ""
         - name: RELATED_IMAGE_TENSORFLOW
-          value: ""          
+          value: ""
         - name: RELATED_IMAGE_EXPLAINER
-          value: ""          
+          value: ""
         - name: RELATED_IMAGE_MOCK_CLASSIFIER
-          value: ""          
-        - name: MANAGER_CREATE_RESOURCES          
+          value: ""
+        - name: MANAGER_CREATE_RESOURCES
           value: "false"
         - name: POD_NAMESPACE
           valueFrom:
@@ -100,7 +101,7 @@ spec:
         - name: EXECUTOR_SERVER_PORT
           value: "8000"
         - name: EXECUTOR_CONTAINER_USER
-          value: "8888"          
+          value: "8888"
         - name: EXECUTOR_CONTAINER_SERVICE_ACCOUNT_NAME
           value: default
         - name: EXECUTOR_SERVER_METRICS_PORT_NAME

diff --git a/operator/helm/split_resources.py b/operator/helm/split_resources.py
@@ -211,6 +211,17 @@ def helm_namespace_override():
                     '{{- if .Values.singleNamespace }}--namespace={{ include "seldon.namespace" . }}{{- end }}'
                 )
 
+                # Update metrics port
+                res["spec"]["template"]["metadata"]["annotations"]["prometheus.io/port"] = helm_value('metrics.port')
+                for portSpec in res["spec"]["template"]["spec"]["containers"][0][
+                    "ports"
+                ]:
+                    if portSpec["name"] == "metrics":
+                        portSpec["containerPort"] = helm_value("metrics.port")
+
+                # Networking
+                res["spec"]["template"]["spec"]["hostNetwork"] = helm_value("hostNetwork")
+
             if kind == "configmap" and name == "seldon-config":
                 res["data"]["credentials"] = helm_value_json("credentials")
                 res["data"]["predictor_servers"] = helm_value_json(
@@ -454,6 +465,14 @@ def helm_namespace_override():
                     re.M,
                 )
 
+            # make sure hostNetwork is not quoted as its a bool
+            fdata = fdata.replace(
+                "'{{ .Values.hostNetwork }}'", "{{ .Values.hostNetwork }}"
+            )
+            # make sure metrics.port is not quoted as its an int
+            fdata = fdata.replace(
+                "containerPort: '{{ .Values.metrics.port }}'", "containerPort: {{ .Values.metrics.port }}"
+            )
             # make sure webhook is not quoted as its an int
             fdata = fdata.replace(
                 "'{{ .Values.webhook.port }}'", "{{ .Values.webhook.port }}"