DEVOPS-3406 - helm: declared secretsMode with ESO ExternalSecret support

packaging/helm/openwork-ee/README.md

@@ -284,20 +284,86 @@ Published self-host planning pages:
 
 ## Secrets
 
-The chart can create an Opaque Secret from `secret.values`, or consume an existing Secret:
+The deployment declares how it manages secrets with a single key:
 
 ```yaml
 secret:
+  secretsMode: inline # inline | existingSecret | externalSecrets
+```
+
+- `inline` (default): the chart renders an Opaque Secret from `secret.values`.
+  Local evaluation only — the values live wherever the values file lives, so
+  never commit real credentials.
+- `existingSecret`: workloads consume a pre-created Secret named by
+  `secret.existingSecret` (requires `secret.create: false`); the chart renders
+  no secret resource.
+- `externalSecrets`: the chart renders an
+  [External Secrets Operator](https://external-secrets.io/) `ExternalSecret`
+  that materializes the workload Secret from an external provider — the
+  GitOps/ArgoCD-safe path, where git holds only store references and remote
+  key paths (requires `secret.create: false`).
+
+Any `secret.keys` override applies in every mode, since all three resolve the
+workload Secret through the same names. The mode combinations are enforced at
+render time: an unknown `secretsMode`, a missing `secret.existingSecret` in
+`existingSecret` mode, `secret.existingSecret` set in any other mode, and
+`secret.create: true` outside `inline` mode all fail the render.
+
+### existingSecret
+
+```yaml
+secret:
+  secretsMode: existingSecret
   create: false
   existingSecret: openwork-ee-secrets
 ```
 
-The existing Secret must contain the keys listed under `secret.keys`, especially:
+The existing Secret must be created in the `namespace` where the chart deploys
+(default `openwork`) before the workloads start, and must contain the keys
+listed under `secret.keys`, especially:
 
 - `DATABASE_URL`
 - `BETTER_AUTH_SECRET`
 - `DEN_DB_ENCRYPTION_KEY`
 
+### externalSecrets (GitOps / ArgoCD)
+
+For GitOps flows (ArgoCD runs `helm template`, so anything in values lands in
+git and in rendered manifests), use ESO mode. The chart renders an
+`ExternalSecret` that materializes the same-named workload Secret from your
+provider in-cluster:
+
+The chart renders `spec.data` — the oldest stable ESO shape, unchanged since
+`external-secrets.io/v1beta1` — pulling every `secret.keys.*` entry from
+`<pathPrefix>/<KEY_NAME>` in the provider. The key list is generated from
+`secret.keys`, so it can never drift from what the workloads consume:
+
+```yaml
+secret:
+  secretsMode: externalSecrets
+  create: false
+externalSecrets:
+  secretStoreRef:
+    # References an existing (Cluster)SecretStore; for AWS Secrets Manager the
+    # store itself carries spec.provider.aws (region, auth), the chart only
+    # points at it by name.
+    name: external-secrets
+    kind: ClusterSecretStore
+  refreshInterval: 5m
+  # Every secret.keys.* value must exist as a JSON property under this trunk,
+  # e.g. eks/openwork/prod/den/DATABASE_URL.
+  pathPrefix: "eks/openwork/prod/den"
+```
+
+Every property the workloads consume must exist in your provider under
+`pathPrefix`, named like `secret.keys.*` values (`DATABASE_URL`,
+`BETTER_AUTH_SECRET`, ...) — the chart pulls each key by name and cannot
+invent missing ones. `target.deletionPolicy` defaults to `Retain`, so
+uninstalling the release keeps the materialized Secret. ESO must be installed
+on the destination cluster with a `SecretStore`/`ClusterSecretStore`; the
+chart selects `external-secrets.io/v1` or `v1beta1` from cluster capabilities
+and fails loudly at sync time if the CRDs are missing.
+
 Set optional `DATABASE_REDIS_URL` to enable Den API Redis-backed session and query caching. Set `DAYTONA_API_KEY` when `config.provisioner.mode` is `daytona`. Set `POLAR_ACCESS_TOKEN` when Polar feature gating is enabled. Set `OPENROUTER_MANAGEMENT_API_KEY` when enabling OpenWork Models management.
 
 Redis cache examples:

packaging/helm/openwork-ee/templates/_helpers.tpl

@@ -7,7 +7,13 @@
 {{- .Values.fullnameOverride | trunc 63 | trimSuffix "-" -}}
 {{- else -}}
 {{- $name := default .Chart.Name .Values.nameOverride -}}
-{{- if contains $name .Release.Name -}}
+{{- /*
+  Dedupe in both directions: contains handles release names that already
+  include the chart name (my-openwork-ee); hasPrefix handles release names
+  that prefix the chart name (release "openwork", chart "openwork-ee"), which
+  would otherwise produce doubled names like openwork-openwork-ee-secret.
+*/ -}}
+{{- if or (contains $name .Release.Name) (hasPrefix .Release.Name $name) -}}
 {{- .Release.Name | trunc 63 | trimSuffix "-" -}}
 {{- else -}}
 {{- printf "%s-%s" .Release.Name $name | trunc 63 | trimSuffix "-" -}}
@@ -59,11 +65,101 @@ app.kubernetes.io/component: {{ .component }}
 {{- end -}}
 {{- end -}}
 
+{{/*
+  Returns the workload Secret name as a quoted string: existingSecret values
+  are user-supplied and may look like YAML scalars (true, 1234), which would
+  otherwise render non-string manifest fields and fail at apply time.
+  Consumers that need the bare name trim the quotes.
+*/}}
 {{- define "openwork-ee.secretName" -}}
+{{- if eq .Values.secret.secretsMode "existingSecret" -}}
+{{- .Values.secret.existingSecret | toString | trim | quote -}}
+{{- else -}}
+{{- printf "%s-secret" (include "openwork-ee.fullname" .) | quote -}}
+{{- end -}}
+{{- end -}}
+
+{{/* Bare (unquoted) Secret name for contexts that need it (e.g. lookup). */}}
+{{- define "openwork-ee.secretNameRaw" -}}
+{{- include "openwork-ee.secretName" . | trimAll "\"" -}}
+{{- end -}}
+
+{{/* Bare (unquoted) namespace name for contexts that need it. */}}
+{{- define "openwork-ee.namespaceRaw" -}}
+{{- include "openwork-ee.namespace" . | trimAll "\"" -}}
+{{- end -}}
+
+{{- define "openwork-ee.secretsMode.validate" -}}
+{{- if not (has .Values.secret.secretsMode (list "inline" "existingSecret" "externalSecrets")) -}}
+{{- fail "secretsMode must be one of inline, existingSecret, externalSecrets" -}}
+{{- end -}}
+{{- if eq .Values.secret.secretsMode "existingSecret" -}}
+{{- if not (.Values.secret.existingSecret | toString | trim) -}}
+{{- fail "secret.existingSecret is required when secretsMode=existingSecret" -}}
+{{- end -}}
+{{- end -}}
+{{- if ne .Values.secret.secretsMode "existingSecret" -}}
 {{- if .Values.secret.existingSecret -}}
-{{- .Values.secret.existingSecret -}}
+{{- fail "secret.existingSecret is only allowed when secretsMode=existingSecret" -}}
+{{- end -}}
+{{- end -}}
+{{- if ne .Values.secret.secretsMode "inline" -}}
+{{- if .Values.secret.create -}}
+{{- fail "secret.create must be false when secretsMode is not inline" -}}
+{{- end -}}
+{{- end -}}
+{{- if and (eq .Values.secret.secretsMode "inline") (not .Values.secret.create) -}}
+{{- /* Legacy migration shim: values files from before secretsMode shipped that
+       set create=false with untouched placeholder values meant "no inline
+       secrets" — treat that as existingSecret mode. Real-looking values with
+       create=false are incoherent and must fail, not be silently rerouted. */ -}}
+{{- $dsn := .Values.secret.values.databaseUrl | toString -}}
+{{- $auth := .Values.secret.values.betterAuthSecret | toString -}}
+{{- $enc := .Values.secret.values.denDbEncryptionKey | toString -}}
+{{- /*
+  Reroute only when ALL three required values are still placeholders: a
+  partially-filled inline block means someone set real values and create=false
+  is incoherent — fail rather than silently ignoring their real values.
+*/ -}}
+{{- $dsnIsPlaceholder := or (contains "change-me@" $dsn) (contains "******" $dsn) -}}
+{{- if and (hasPrefix "CHANGE_ME" $auth) (hasPrefix "CHANGE_ME" $enc) $dsnIsPlaceholder -}}
+{{- $_ := set .Values.secret "secretsMode" "existingSecret" -}}
+{{- if not (.Values.secret.existingSecret | toString | trim) -}}
+{{- $_ := set .Values.secret "existingSecret" (include "openwork-ee.fullname" . | printf "%s-secret") -}}
+{{- end -}}
 {{- else -}}
-{{- include "openwork-ee.fullname" . }}-secret
+{{- fail "secret.create must be true when secretsMode=inline (set secretsMode=existingSecret or externalSecrets to source secrets externally)" -}}
+{{- end -}}
+{{- end -}}
+{{- end -}}
+
+{{- define "openwork-ee.externalSecrets.apiVersion" -}}
+{{- if .Capabilities.APIVersions.Has "external-secrets.io/v1" -}}
+external-secrets.io/v1
+{{- else -}}
+external-secrets.io/v1beta1
+{{- end -}}
+{{- end -}}
+
+{{- define "openwork-ee.externalSecrets.validate" -}}
+{{- if eq .Values.secret.secretsMode "externalSecrets" -}}
+{{- $storeName := "" -}}
+{{- if .Values.externalSecrets.secretStoreRef -}}
+{{- $storeName = .Values.externalSecrets.secretStoreRef.name | toString | trim -}}
+{{- end -}}
+{{- if not $storeName -}}
+{{- fail "externalSecrets.secretStoreRef.name is required when secretsMode=externalSecrets" -}}
+{{- end -}}
+{{- $storeKind := "" -}}
+{{- if .Values.externalSecrets.secretStoreRef -}}
+{{- $storeKind = .Values.externalSecrets.secretStoreRef.kind | toString -}}
+{{- end -}}
+{{- if not (has $storeKind (list "SecretStore" "ClusterSecretStore")) -}}
+{{- fail "externalSecrets.secretStoreRef.kind must be SecretStore or ClusterSecretStore" -}}
+{{- end -}}
+{{- if not (.Values.externalSecrets.pathPrefix | toString | trim) -}}
+{{- fail "externalSecrets.pathPrefix is required when secretsMode=externalSecrets" -}}
+{{- end -}}
 {{- end -}}
 {{- end -}}
 

packaging/helm/openwork-ee/templates/configmap.yaml

@@ -1,3 +1,4 @@
+{{- include "openwork-ee.secretsMode.validate" . }}
 apiVersion: v1
 kind: ConfigMap
 metadata:

packaging/helm/openwork-ee/templates/externalsecret.yaml

@@ -0,0 +1,43 @@
+{{- include "openwork-ee.secretsMode.validate" . }}
+{{- if eq .Values.secret.secretsMode "externalSecrets" }}
+{{- include "openwork-ee.externalSecrets.validate" . }}
+apiVersion: {{ include "openwork-ee.externalSecrets.apiVersion" . }}
+kind: ExternalSecret
+metadata:
+  name: {{ include "openwork-ee.secretName" . }}
+  namespace: {{ include "openwork-ee.namespace" . }}
+  labels:
+    {{- include "openwork-ee.labels" . | nindent 4 }}
+  {{- /*
+    When the migration Job runs as a pre-install/pre-upgrade hook, the
+    ExternalSecret must apply first so ESO can start materializing the Secret
+    the Job consumes. Weight -10 precedes the Job's -5. ArgoCD maps these to
+    PreSync the same way, so GitOps installs get the same ordering.
+  */}}
+  {{- if .Values.migrations.hook }}
+  annotations:
+    "helm.sh/hook": pre-install,pre-upgrade
+    "helm.sh/hook-weight": "-10"
+    "helm.sh/hook-delete-policy": before-hook-creation
+  {{- end }}
+spec:
+  refreshInterval: {{ .Values.externalSecrets.refreshInterval | quote }}
+  secretStoreRef:
+    name: {{ .Values.externalSecrets.secretStoreRef.name | toString | trim | quote }}
+    kind: {{ .Values.externalSecrets.secretStoreRef.kind }}
+  target:
+    name: {{ include "openwork-ee.secretName" . }}
+    creationPolicy: {{ .Values.externalSecrets.target.creationPolicy | default "Owner" }}
+    deletionPolicy: {{ .Values.externalSecrets.target.deletionPolicy | default "Retain" }}
+  data:
+    {{- $prefix := .Values.externalSecrets.pathPrefix | toString | trim | trimSuffix "/" }}
+    {{- range $name := keys .Values.secret.keys | sortAlpha }}
+    {{- $envKey := index $.Values.secret.keys $name }}
+    - secretKey: {{ $envKey | quote }}
+      remoteRef:
+        key: {{ printf "%s/%s" $prefix $envKey | quote }}
+        conversionStrategy: {{ $.Values.externalSecrets.conversionStrategy }}
+        decodingStrategy: {{ $.Values.externalSecrets.decodingStrategy }}
+        metadataPolicy: {{ $.Values.externalSecrets.metadataPolicy }}
+    {{- end }}
+{{- end }}

packaging/helm/openwork-ee/templates/migration-job.yaml

@@ -1,4 +1,6 @@
 {{- include "openwork-ee.customCa.validate" . }}
+{{- /* Runs the legacy create=false shim before any secretsMode check below. */}}
+{{- include "openwork-ee.secretsMode.validate" . }}
 {{- if .Values.migrations.enabled }}
 apiVersion: batch/v1
 kind: Job
@@ -28,6 +30,9 @@ spec:
       {{- end }}
     spec:
       restartPolicy: Never
+      {{- if ne .Values.secret.secretsMode "inline" }}
+      serviceAccountName: {{ include "openwork-ee.fullname" . }}-migrate
+      {{- end }}
       {{- with .Values.imagePullSecrets }}
       imagePullSecrets:
         {{- toYaml . | nindent 8 }}
@@ -36,6 +41,26 @@ spec:
       volumes:
         {{- include "openwork-ee.customCa.volume" . | nindent 8 }}
       {{- end }}
+      {{- /*
+        When the Secret is materialized asynchronously (ESO or out-of-band
+        creation), the Job must not fail on a missing Secret before it exists.
+        Block in an init container until it appears instead of erroring on the
+        env secretKeyRef.
+      */}}
+      {{- if ne .Values.secret.secretsMode "inline" }}
+      initContainers:
+        - name: wait-for-secret
+          image: "{{ .Values.migrations.kubectlImage.repository }}:{{ .Values.migrations.kubectlImage.tag }}"
+          imagePullPolicy: {{ .Values.image.pullPolicy }}
+          command:
+            - sh
+            - -c
+            - |
+              until kubectl get secret {{ include "openwork-ee.secretNameRaw" . }} -n {{ include "openwork-ee.namespaceRaw" . }} > /dev/null 2>&1; do
+                echo "waiting for secret {{ include "openwork-ee.secretNameRaw" . }}..."
+                sleep 3
+              done
+      {{- end }}
       containers:
         - name: migrate
           image: "{{ .Values.denApi.image.repository }}:{{ default .Values.image.tag .Values.denApi.image.tag }}"
@@ -52,22 +77,22 @@ spec:
             - name: DB_MODE
               value: {{ .Values.config.databaseMode | quote }}
             - name: DATABASE_URL
-              {{- if .Values.secret.create }}
+              {{- if eq .Values.secret.secretsMode "inline" }}
               value: {{ .Values.secret.values.databaseUrl | quote }}
               {{- else }}
               valueFrom:
                 secretKeyRef:
                   name: {{ include "openwork-ee.secretName" . }}
-                  key: {{ .Values.secret.keys.databaseUrl }}
+                  key: {{ .Values.secret.keys.databaseUrl | quote }}
               {{- end }}
             - name: DEN_DB_ENCRYPTION_KEY
-              {{- if .Values.secret.create }}
+              {{- if eq .Values.secret.secretsMode "inline" }}
               value: {{ .Values.secret.values.denDbEncryptionKey | quote }}
               {{- else }}
               valueFrom:
                 secretKeyRef:
                   name: {{ include "openwork-ee.secretName" . }}
-                  key: {{ .Values.secret.keys.denDbEncryptionKey }}
+                  key: {{ .Values.secret.keys.denDbEncryptionKey | quote }}
               {{- end }}
             {{- if .Values.customCa.enabled }}
             {{- include "openwork-ee.customCa.env" . | nindent 12 }}

packaging/helm/openwork-ee/templates/migration-rbac.yaml

@@ -0,0 +1,66 @@
+{{- /*
+  RBAC for the migration Job's wait-for-secret init container: the Job polls
+  for an asynchronously materialized Secret (ESO / out-of-band creation)
+  instead of failing on the env secretKeyRef. Rendered only outside inline
+  mode. In hook mode these must be hooks with an earlier weight than the Job
+  (-6 precedes the Job's -5) so they exist before it starts.
+*/}}
+{{- /* Runs the legacy create=false shim before the secretsMode check below. */}}
+{{- include "openwork-ee.secretsMode.validate" . }}
+{{- if and .Values.migrations.enabled (ne .Values.secret.secretsMode "inline") }}
+{{- $asHook := .Values.migrations.hook }}
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: {{ include "openwork-ee.fullname" . }}-migrate
+  namespace: {{ include "openwork-ee.namespace" . }}
+  labels:
+    {{- include "openwork-ee.componentLabels" (dict "root" . "component" "migration") | nindent 4 }}
+  {{- if $asHook }}
+  annotations:
+    "helm.sh/hook": pre-install,pre-upgrade
+    "helm.sh/hook-weight": "-6"
+    "helm.sh/hook-delete-policy": before-hook-creation
+  {{- end }}
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: Role
+metadata:
+  name: {{ include "openwork-ee.fullname" . }}-migrate
+  namespace: {{ include "openwork-ee.namespace" . }}
+  labels:
+    {{- include "openwork-ee.componentLabels" (dict "root" . "component" "migration") | nindent 4 }}
+  {{- if $asHook }}
+  annotations:
+    "helm.sh/hook": pre-install,pre-upgrade
+    "helm.sh/hook-weight": "-6"
+    "helm.sh/hook-delete-policy": before-hook-creation
+  {{- end }}
+rules:
+  - apiGroups: [""]
+    resources: ["secrets"]
+    resourceNames: [{{ include "openwork-ee.secretName" . }}]
+    verbs: ["get"]
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: RoleBinding
+metadata:
+  name: {{ include "openwork-ee.fullname" . }}-migrate
+  namespace: {{ include "openwork-ee.namespace" . }}
+  labels:
+    {{- include "openwork-ee.componentLabels" (dict "root" . "component" "migration") | nindent 4 }}
+  {{- if $asHook }}
+  annotations:
+    "helm.sh/hook": pre-install,pre-upgrade
+    "helm.sh/hook-weight": "-6"
+    "helm.sh/hook-delete-policy": before-hook-creation
+  {{- end }}
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: Role
+  name: {{ include "openwork-ee.fullname" . }}-migrate
+subjects:
+  - kind: ServiceAccount
+    name: {{ include "openwork-ee.fullname" . }}-migrate
+    namespace: {{ include "openwork-ee.namespace" . }}
+{{- end }}