fix: csp monitor fixes

Signed-off-by: Tanisha goyal <[email protected]>

Author: tanishagoyal2

distros/kubernetes/nvsentinel/charts/csp-health-monitor/templates/deployment.yaml

@@ -55,7 +55,6 @@ spec:
         - name: {{ .Chart.Name }}
           image: "{{ .Values.image.repository }}:{{ .Values.image.tag | default ((.Values.global).image).tag | default .Chart.AppVersion }}"
           imagePullPolicy: {{ .Values.image.pullPolicy }}
-          command: ["/app/csp-health-monitor"]
           args:
           - "--config=/etc/config/config.toml"
           - "--metrics-port={{ ((.Values.global).metricsPort) | default 2112 }}"
@@ -81,11 +80,10 @@ spec:
                 optional: true
 
         - name: maintenance-notifier
-          image: "{{ .Values.image.repository }}:{{ .Values.image.tag | default ((.Values.global).image).tag | default .Chart.AppVersion }}"
-          imagePullPolicy: {{ .Values.image.pullPolicy }}
+          image: "{{ .Values.quarantineTriggerEngine.image.repository }}:{{ .Values.image.tag | default ((.Values.global).image).tag | default .Chart.AppVersion }}"
+          imagePullPolicy: {{ .Values.quarantineTriggerEngine.image.pullPolicy | default .Values.image.pullPolicy }}
           securityContext:
             runAsUser: 0
-          command: ["/app/maintenance-notifier"]
           args:
           - "--config=/etc/config/config.toml"
           - "--mongo-client-cert-mount-path=/etc/ssl/mongo-client"

distros/kubernetes/nvsentinel/charts/csp-health-monitor/templates/serviceaccount.yaml

@@ -23,6 +23,6 @@ metadata:
     iam.gke.io/gcp-service-account: {{ .Values.configToml.gcp.gcpServiceAccountName }}@{{ .Values.configToml.gcp.targetProjectId }}.iam.gserviceaccount.com
       {{- end }}
       {{- if and (eq .Values.cspName "aws") .Values.configToml.aws .Values.configToml.aws.accountId .Values.configToml.clusterName }}
-    eks.amazonaws.com/role-arn: arn:aws:iam::{{ .Values.configToml.aws.accountId }}:role/{{ .Values.configToml.clusterName }}-health-monitor
+    eks.amazonaws.com/role-arn: arn:aws:iam::{{ .Values.configToml.aws.accountId }}:role/{{ .Values.configToml.clusterName }}-nvsentinel-health-monitor-assume-role-policy
       {{- end }}
-    {{- end }}
+    {{- end }}

distros/kubernetes/nvsentinel/charts/csp-health-monitor/values.yaml

@@ -29,6 +29,9 @@ resources:
 
 # Configuration for the Quarantine Trigger Engine sidecar
 quarantineTriggerEngine:
+  image:
+    repository: ghcr.io/nvidia/nvsentinel/maintenance-notifier
+    pullPolicy: IfNotPresent
   resources:
     limits:
       cpu: "500m"