Add hardened Hub API security contexts and writable scratch (0.170.1)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

.github/workflows/http-port-migration.yaml

@@ -0,0 +1,28 @@
+name: HTTP port migration
+
+on:
+  pull_request:
+  workflow_dispatch:
+
+jobs:
+  migration:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: azure/setup-helm@v4.3.1
+      - uses: helm/kind-action@v1
+      - name: Validate migration against Kubernetes
+        env:
+          SERVER_VALIDATION: "true"
+        run: bash scripts/check-http-port-migration.sh
+      - name: Validate chart runtime compatibility
+        run: |
+          helm lint charts/hub
+          bash scripts/check-admin-runtime.sh
+          bash scripts/check-hardened-workers.sh
+          bash scripts/check-frontend-runtime.sh
+          bash scripts/check-hub-api-runtime.sh
+      - name: Smoke test hardened frontend images and Service routing
+        env:
+          CLUSTER_SMOKE: "true"
+        run: bash scripts/check-frontend-runtime.sh

charts/hub/Chart.yaml

@@ -16,7 +16,7 @@ type: application
 # This is the chart version. This version number should be incremented each time you make changes
 # to the chart and its templates, including the app version.
 # Versions are expected to follow Semantic Versioning (https://semver.org/)
-version: 0.168.2
+version: 0.170.1
 
 # This is the version number of the application being deployed. This version number should be
 # incremented each time you make changes to the application. Versions are not expected to

charts/hub/README.md

@@ -31,12 +31,41 @@ Uninstall the Kerberos Hub chart
 
     helm uninstall hub -n kerberos-hub
 
+### Hardened Hub API
+
+Starting with chart **`0.170.1`**, `kerberoshub.api.securityContext` is rendered
+on the Hub API container and `kerberoshub.api.podSecurityContext` on its pod.
+The current image runs as `65532:65532` with a read-only root filesystem,
+privilege escalation disabled, all capabilities dropped and `RuntimeDefault`
+seccomp. A disk-backed `hub-api-tmp` emptyDir mounted at `/tmp`, with pod
+`fsGroup: 65532`, keeps attachment uploads and multipart parsing writable.
+API Service, ingress and port settings are unchanged.
+
+Append custom mounts and volumes to the default lists rather than replacing
+`hub-api-tmp`. Helm replaces lists in overrides. Existing MongoDB TLS, API TLS,
+catalog and workflow-configuration mounts remain supported.
+
+For a legacy image that requires its original identity or writable root
+filesystem, explicitly set both `kerberoshub.api.securityContext: null` and
+`kerberoshub.api.podSecurityContext: null`. Custom security settings are
+per-API; other workloads do not inherit them. When upgrading, check retained
+values (`--reuse-values` or GitOps overrides) so the security context and
+writable scratch mount stay aligned.
+
+Run `bash scripts/check-hub-api-runtime.sh` to validate defaults, overrides,
+legacy opt-out, TLS/configuration mounts and deployment modes.
+
 ### Hardened frontend images
 
-New shell-free frontend images run as `65532:65532` and listen on port `8080`.
-Before selecting such an image tag, set the following values together. Defaults
-remain compatible with legacy images on port `80`; this chart does not upgrade
-the frontend image automatically. The Service and ingress remain on port `80`.
+Starting with chart **`0.170.0`**, both frontends default to shell-free
+`ghcr.io/uug-ai/hub-frontend:v1.19.28`, running as `65532:65532` on port `8080`
+with a read-only root filesystem and writable `/tmp`. The Service and ingress
+remain on port `80`. Admin retains its legacy image and port `80`; its hardened
+runtime remains opt-in.
+
+The following frontend runtime settings are now the defaults. When upgrading,
+remove stale port `80` overrides and coordinate image/runtime values; Helm
+`--reuse-values` and GitOps overrides may retain the old settings.
 Use chart `0.167.2` or newer when the demo frontend is enabled; `0.167.1`
 supports these settings only on the primary frontend.
 
@@ -73,6 +102,26 @@ the primary and demo frontends, which share the frontend image tag. Update the
 live Argo CD Application's chart pin as well as its values: an image-only
 upgrade leaves older charts forwarding to port `80` and causes gateway errors.
 
+For legacy frontend images or a rollback, explicitly restore **all** runtime
+settings together (including removing inherited security defaults):
+
+```yaml
+kerberoshub:
+    frontend:
+        tag: v1.13.9
+        containerPort: 80
+        securityContext: null
+        volumes: []
+        volumeMounts: []
+```
+
+Keep any application-specific branding mounts when adjusting these lists.
+Chart upgrades do not rewrite separately managed NodePort Services: ensure they
+target the named `http` port, not numeric `80`. Follow the duplicate-port recovery
+instructions below if the existing Deployment cannot be patched to `8080`.
+Run `bash scripts/check-frontend-runtime.sh` for default, demo, legacy override,
+and deployment-values regression coverage.
+
 ### Hardened Admin image (opt-in, chart 0.167.3)
 
 `charts/hub` deploys Admin in `all` and `ui` modes. Shell-free Admin `v1.4.3`
@@ -128,6 +177,57 @@ security context, and disable `/healthz` probes together. Run
 `bash scripts/check-admin-runtime.sh` from the repository root to verify
 legacy defaults, hardened runtime/probes, and unchanged OAuth2 routing.
 
+### Recovering a duplicate HTTP port during upgrades
+
+Chart `0.169.1` publishes this recovery guidance. The helper is available in the
+repository's `scripts/` directory, not in the chart archive. This release does
+not automatically migrate live container ports or change runtime defaults.
+
+An upgrade changing a container's HTTP port from `80` to `8080` can fail with
+`spec.template.spec.containers[0].ports[1].name: Duplicate value: "http"`.
+Kubernetes strategic merges identify container ports by number, not name:
+depending on the live state and apply mechanism, the old `80` entry can survive
+beside the new `8080` entry. Fresh chart renders contain only one HTTP port.
+This is not a reason to rename `http`: Services and probes reference that name.
+
+Inspect the failing Deployment and confirm the image's listening port first:
+
+```sh
+kubectl -n kerberos-hub get deployment hub-frontend -o yaml
+```
+
+In a checkout of this repository, use the targeted migration helper to replace
+the named HTTP entry (not strategically merge it). It supports `admin`,
+`hub-frontend` and `hub-frontend-demo`, preserves unrelated ports and containers,
+and rejects unfamiliar HTTP configurations. It requires `kubectl` and `jq` and
+uses the current kubeconfig context. The default is a **server dry run**:
+
+```sh
+bash scripts/migrate-http-port.sh kerberos-hub hub-frontend 8080
+# Apply only after inspecting the dry run and confirming the target/context.
+bash scripts/migrate-http-port.sh kerberos-hub hub-frontend 8080 --apply
+```
+
+Run this only for affected Deployments, using the port selected in your upgrade
+values. Then retry the Helm upgrade or GitOps sync with the coordinated image,
+port, security context and scratch-volume settings above. Pause GitOps auto-sync
+during manual recovery so it cannot race or undo the patch, and resume it after
+the desired state has synced. The helper uses resource-version/name tests to
+reject concurrent changes; re-inspect and retry if those tests fail.
+
+Changing the pod template triggers a rollout, and setting `8080` while the old
+image still listens on `80` can temporarily interrupt Service traffic. Schedule
+the recovery in a maintenance window and complete the matching image upgrade
+immediately. The helper does not change images, mount scratch volumes, or repair
+unrelated readiness failures. For a legacy-image rollback use port `80` instead,
+restore the matching runtime settings, and retry the rollback. Do not use blanket
+`helm upgrade --force` or resource replacement for unrelated workloads.
+
+Run `bash scripts/check-http-port-migration.sh` to reproduce the strategic-merge
+failure and test targeted replacement, idempotency, and reverse migration for all
+three Deployments. CI also checks API rejection of duplicate names and acceptance
+of the repaired spec against a disposable Kind cluster.
+
 ### Hardened Hub workers (opt-in, chart 0.167.3)
 
 The worker templates accept **per-component** `securityContext` (container) and
@@ -614,6 +714,7 @@ kerberospipeline:
 | `kerberoshub.api.mfaIssuer` | When enabling the MFA access, this is the name that will be shown in the MFA app. | `"yourdomain.com"` |
 | `kerberoshub.api.apiKey` | API key for `kerberoshub.api`. | `"a-random-admin-api-key"` |
 | `kerberoshub.api.defaultTaskRetentionDays` | Default retention (in days) applied to tasks without an explicit `retention_days`. New tasks are stamped with this value. Set to `"0"` or a negative value to keep tasks indefinitely. Must match `kerberoshub.cleanup.defaultTaskRetentionDays`. | `"0"` |
+| `kerberoshub.api.auditExportMaxEventsPerFile` | Most audit events one export file holds (`AUDIT_EXPORT_MAX_EVENTS_PER_FILE`). It does not refuse larger exports: the Audit page downloads them as several files of at most this size. An integer from 1 to 1000000; empty uses the API default of 100000. | `""` |
 | `kerberoshub.api.tls` | Bring your own TLS certificates for Kerberos Hub API ingress. | `<list>` |
 | `kerberoshub.api.tls.secretName` | Kubernetes Secret name used by `kerberoshub.api.tls`. | `""` |
 | `kerberoshub.api.language` | The language of Kerberos Hub API responses, error messages will be communicated in the specified language. | `"english"` |
@@ -795,12 +896,12 @@ kerberospipeline:
 | `kerberoshub.cleanup.resources.requests.cpu` | CPU request for `kerberoshub.cleanup`. | `"10m"` |
 | `kerberoshub.audit.repository` | Hub audit service container image repository. | `"ghcr.io/uug-ai/hub-audit"` |
 | `kerberoshub.audit.pullPolicy` | Hub audit service image pull policy. | `"IfNotPresent"` |
-| `kerberoshub.audit.tag` | Hub audit service image tag. | `"v1.0.1"` |
+| `kerberoshub.audit.tag` | Hub audit service image tag. S3 destinations need v1.0.9 or later, and `s3.allowUnlockedBucket` needs v1.0.10. | `"v1.0.10"` |
 | `kerberoshub.audit.replicas` | Number of audit service replicas. Per-destination MongoDB leases prevent concurrent delivery. | `1` |
 | `kerberoshub.audit.dispatchInterval` | How often the mounted destination configuration is reloaded and eligible destinations are polled. | `"5s"` |
 | `kerberoshub.audit.leaseDuration` | Per-destination lease duration. It must exceed every destination timeout by at least 30 seconds. | `"2m"` |
 | `kerberoshub.audit.terminationGracePeriodSeconds` | Pod termination grace period. Keep this longer than `leaseDuration` so an in-flight cycle can finish. | `150` |
-| `kerberoshub.audit.destinations` | Audit destinations. Every entry supports `id`, `type` (`webhook` by default, or `s3`), `enabled`, delivery limits and filters. Webhooks add `url`, public `headers`, `bearerTokenSecret` and arbitrary `secretHeaders`. S3 destinations add `s3.bucket`, `s3.region`, optional `s3.prefix`, `s3.endpoint` and `s3.forcePathStyle` for S3-compatible stores, and an optional `credentialsSecret` (`name`, `accessKeyIdKey`, `secretAccessKeyKey`); without it the pod's AWS identity is used. The bucket must have Object Lock with a `COMPLIANCE` default retention unless `s3.allowUnlockedBucket` is `true` (stores without Object Lock such as Garage; the archive is then deletable). Destination IDs retain independent checkpoints. | `[]` |
+| `kerberoshub.audit.destinations` | Audit destinations. Every entry supports `id`, `type` (`webhook` by default, or `s3`), `enabled`, delivery limits and filters. Webhooks add `url`, public `headers`, `bearerTokenSecret` and arbitrary `secretHeaders`. S3 destinations add `s3.bucket`, `s3.region`, optional `s3.prefix`, `s3.endpoint` and `s3.forcePathStyle` for S3-compatible stores, and an optional `credentialsSecret` (`name`, `accessKeyIdKey`, `secretAccessKeyKey`); without it the pod's AWS identity is used. The bucket must have Object Lock with a `COMPLIANCE` default retention unless `s3.allowUnlockedBucket` is `true` (stores without Object Lock such as Garage; the archive is then deletable). Every field, the delivery semantics and per-provider S3 settings are documented at https://learn.uug.ai/docs/hub/audit-events/#export-audit-events. Destination IDs retain independent checkpoints. | `[]` |
 | `kerberoshub.audit.serviceAccount.create` | Create a ServiceAccount used only by hub-audit, for example for S3 access through IRSA. | `false` |
 | `kerberoshub.audit.serviceAccount.name` | Name of the hub-audit ServiceAccount; with `create: false`, an existing ServiceAccount to use. Defaults to `hub-audit` when created; when empty and not created, `kerberoshub.serviceAccount` applies. | `""` |
 | `kerberoshub.audit.serviceAccount.annotations` | Annotations for the hub-audit ServiceAccount, such as `eks.amazonaws.com/role-arn`. | `{}` |
@@ -815,7 +916,7 @@ kerberospipeline:
 | `kerberoshub.monitordevice.tag` | The Docker image tag/version. | `"v1.4.0"` |
 | `kerberoshub.monitordevice.replicas` | Number of replicas for `kerberoshub.monitordevice`. | `1` |
 | `kerberoshub.monitordevice.logLevel` | Log verbosity level for `kerberoshub.monitordevice`. | `"info"` |
-| `kerberoshub.monitordevice.scanIntervalSeconds` | Seconds between device status scans. Sets how quickly a camera going offline or online is detected and, for accounts with "Group notifications" on, how long notifications are held to collect other cameras that change status at the same time. | `60` |
+| `kerberoshub.monitordevice.scanIntervalSeconds` | Seconds between device status scans. Sets how quickly a camera going offline or online is detected. For accounts with "Group notifications" on, `hub-monitor-device` v1.4.10 and later then hold a change 20 seconds (`DEVICE_GROUPING_HOLD_SECONDS`) to collect other cameras that change status at the same time; older releases hold it one scan. | `60` |
 | `kerberoshub.monitordevice.resources.requests.memory` | Memory request for `kerberoshub.monitordevice`. | `"10Mi"` |
 | `kerberoshub.monitordevice.resources.requests.cpu` | CPU request for `kerberoshub.monitordevice`. | `"10m"` |
 | `kerberoshub.reactivate.repository` | Container image repository for `kerberoshub.reactivate`. | `"uugai/hub-reactivatesubscriptions"` |
@@ -1017,22 +1118,32 @@ Sent by `hub-monitor-device` when a camera goes online, offline or idle. The sub
 |---|---|
 | `{{user}}` | Username of the account the camera belongs to. |
 | `{{title}}` | The email subject. |
-| `{{text}}` | The status change, for example "`<camera>` is online, but the camera stream is not connected." |
+| `{{text}}` | The change, worded from the previous status (see below). |
 | `{{status}}` | `active` (online), `inactive` (offline) or `idle` (online, camera stream not connected). |
 | `{{devicename}}`, `{{deviceid}}` | Camera name and camera key. |
 | `{{eventdate}}`, `{{eventtime}}`, `{{eventdatetime}}` | When the status changed, in the server's timezone (normally UTC), not the account's. |
 | `{{link}}` | Not set; replaced with an empty value. |
 
+`{{text}}` per change (webhooks and IFTTT get the same text):
+
+| Change | `{{text}}` |
+|---|---|
+| Goes offline | "`<camera>` is offline, it hasn't responded for 5 minutes." |
+| Loses its camera stream while online | "`<camera>` is online, but the camera stream is not connected." |
+| Comes back from offline without its camera stream | "`<camera>` is back online, but the camera stream is not connected." |
+| Gets its camera stream back | "`<camera>`'s camera stream is connected again." |
+| Comes back from offline | "`<camera>` is back online." |
+
 
 #### Grouped camera status: `device_group`
 
-Sent by `hub-monitor-device` instead of `device` when an account has **Group notifications** switched on (Alerts → Devices, stored as `notificationSettings.devices.groupNotifications`) and **several** of its cameras change status in the same scan. A change is held for one scan (`kerberoshub.monitordevice.scanIntervalSeconds`, default 60 seconds) so that cameras of the same outage arrive in one email. When only one camera changed, the usual `device` email is sent. The subject is `email.templates.deviceGroupTitle`; the template name is `email.templates.deviceGroup` (default `device_group`, so custom files are `device_group.html` and `device_group.txt`).
+Sent by `hub-monitor-device` instead of `device` when an account has **Group notifications** switched on (Alerts → Devices, stored as `notificationSettings.devices.groupNotifications`) and **several** of its cameras change status at the same time. Changes are detected on every scan (`kerberoshub.monitordevice.scanIntervalSeconds`, default 60 seconds); a change is then held 20 seconds and the account checked again, so that cameras of the same outage arrive in one email. The hold is the monitor's `DEVICE_GROUPING_HOLD_SECONDS` (default `20`; `0`, or at least the scan interval, holds until the next scan, which is also what releases before v1.4.10 do). When only one camera changed, the usual `device` email is sent. The subject is `email.templates.deviceGroupTitle`; the template name is `email.templates.deviceGroup` (default `device_group`, so custom files are `device_group.html` and `device_group.txt`).
 
 | Placeholder | Value |
 |---|---|
 | `{{user}}` | Name of the account the cameras belong to: first and last name, or the username when both are empty. |
 | `{{title}}` | The email subject. |
-| `{{text}}` | One-line summary, for example "6 cameras changed status: 4 offline, 1 lost their camera stream, 1 back online." |
+| `{{text}}` | One-line summary, for example "6 cameras changed status: 4 offline, 1 lost their camera stream, 1 back online." The other changes read "back online without camera stream" and "camera stream reconnected". |
 | `{{count}}` | Number of cameras in the email. |
 | `{{status}}` | `active`, `inactive` or `idle` when all cameras have the same new status, otherwise `mixed`. |
 | `{{sections}}` | The cameras as HTML, one section per status and site, group or "Other cameras" (see below). |
@@ -1040,7 +1151,7 @@ Sent by `hub-monitor-device` instead of `device` when an account has **Group not
 | `{{link}}` | The cameras page, `<tenant URL>/devices`, built from `kerberoshub.frontend.schema` and `kerberoshub.frontend.url` (see *Links* above). |
 | `{{eventdate}}`, `{{eventtime}}`, `{{eventdatetime}}` | When the email was sent, in the server's timezone (normally UTC). |
 
-**Sections.** Sections are ordered by status (offline, camera stream lost, back online), then sites, groups and "Other cameras". Each camera is listed once: under the site or group that holds the most cameras of the email with the same status (a site wins a tie), or under "Other cameras" when it is in no site or group. A section title reads `Site Gate · Offline · all 3 cameras`, `Group Perimeter · Back online · 2 cameras` or `Other cameras · Camera stream lost · 1 camera`; "all" is added when every camera of the site or group is in the section. Each camera line has its name and a detail: `no response for 12 min`, `agent online, camera stream not connected` or `online`.
+**Sections.** There is one section per change and per site, group or "Other cameras". Sections are ordered by change (offline, camera stream lost, back online without camera stream, camera stream back, back online), then sites, groups and "Other cameras". Each camera is listed once: under the site or group that holds the most cameras of the email with the same change (a site wins a tie), or under "Other cameras" when it is in no site or group. A section title reads `Site Gate · Offline · all 3 cameras`, `Group Perimeter · Back online without camera stream · 2 cameras` or `Other cameras · Camera stream back · 1 camera`; the change labels are `Offline`, `Camera stream lost`, `Back online without camera stream`, `Camera stream back` and `Back online`, and "all" is added when a section of two or more cameras holds every camera of the site or group. Each camera line has its name and a detail: `no response for 12 min`, `agent online, camera stream not connected`, `agent back online, camera stream not connected`, `camera stream connected` or `online`. A camera that changes again during the hold keeps the status from before its first change: one that went offline and came back reads `back online after a brief interruption`, the other reversals `offline again after briefly coming back, no response for 4 min` and `camera stream lost again after briefly reconnecting`.
 
 `{{sections}}` is generated HTML with inline styles (email clients ignore stylesheets): per section a `<p>` with the title, followed by a `<table role="presentation" width="100%">` with one row per camera (name left, detail right). Place it where the camera list should appear; its look can't be changed from the template. For full control over the layout, use `{{sectionsText}}`. Camera, site and group names are HTML-escaped.
 

charts/hub/custom-layout/templates/device_group.html

@@ -276,7 +276,7 @@
                                         letter-spacing: 0em;
                                         text-align: left;
                                         color: #6D6666;
-                                        margin-top: 12px;">Cameras that go offline together usually share a cause, such as a site losing its internet connection or power. Have a look into your Kerberos Hub account for the latest recordings and the status of your cameras.</p>
+                                        margin-top: 12px;">Cameras that change status together usually share a cause, such as a site losing or regaining its internet connection or power. Have a look into your Kerberos Hub account for the latest recordings and the status of your cameras.</p>
 
 						<a style="text-decoration: none;color: none;" href="{{link}}">
                             <p style="font-family: Inter;

charts/hub/templates/kerberos-hub/hub-api.yaml

@@ -140,6 +140,10 @@ spec:
       labels:
         app: hub-api
     spec:
+      {{- with .Values.kerberoshub.api.podSecurityContext }}
+      securityContext:
+        {{- toYaml . | nindent 8 }}
+      {{- end }}
       {{- if .Values.kerberoshub.serviceAccount.create }}
       serviceAccountName: {{ default (printf "%s-%s-sa" .Release.Name .Chart.Name | trunc 63 | trimSuffix "-") .Values.kerberoshub.serviceAccount.name }}
       {{- end }}
@@ -229,6 +233,10 @@ spec:
         - name: hub-api
           image: "{{ .Values.global.imageRegistry }}{{ .Values.kerberoshub.api.repository }}:{{ .Values.kerberoshub.api.tag }}"
           imagePullPolicy: {{ .Values.kerberoshub.api.pullPolicy }}
+          {{- with .Values.kerberoshub.api.securityContext }}
+          securityContext:
+            {{- toYaml . | nindent 12 }}
+          {{- end }}
           {{- with .Values.kerberoshub.api.resources }}
           resources:
             {{- toYaml . | nindent 12 }}
@@ -553,6 +561,10 @@ spec:
               value: "{{ .Values.kerberoshub.api.defaultTaskRetentionDays }}"
             - name: CASES_MAX_RETENTION_DAYS
               value: "{{ .Values.kerberoshub.api.casesMaxRetentionDays }}"
+            {{- with .Values.kerberoshub.api.auditExportMaxEventsPerFile }}
+            - name: AUDIT_EXPORT_MAX_EVENTS_PER_FILE
+              value: {{ . | int64 | toString | quote }}
+            {{- end }}
 
             # SMTP
             - name: SMTP_SERVER

charts/hub/templates/kerberos-hub/hub-frontend-demo.yaml

@@ -110,7 +110,7 @@ spec:
           {{- toYaml . | nindent 12 }}
         {{- end }}
         ports:
-        - containerPort: {{ .Values.kerberoshub.frontend.containerPort | default 80 }}
+        - containerPort: {{ .Values.kerberoshub.frontend.containerPort | default 8080 }}
           name: http
         {{- with .Values.kerberoshub.frontend.volumeMounts}}
         volumeMounts:

charts/hub/templates/kerberos-hub/hub-frontend.yaml

@@ -191,7 +191,7 @@ spec:
           {{- toYaml . | nindent 12 }}
         {{- end }}
         ports:
-        - containerPort: {{ .Values.kerberoshub.frontend.containerPort | default 80 }}
+        - containerPort: {{ .Values.kerberoshub.frontend.containerPort | default 8080 }}
           name: http
         {{- with .Values.kerberoshub.frontend.volumeMounts}}
         volumeMounts:

charts/hub/values.schema.json

@@ -100,6 +100,13 @@
                 "type": "string",
                 "pattern": "^https?://[^/?#]+$"
               }
+            },
+            "auditExportMaxEventsPerFile": {
+              "type": ["integer", "string"],
+              "minimum": 1,
+              "maximum": 1000000,
+              "pattern": "^([1-9][0-9]{0,5}|1000000)?$",
+              "description": "Most audit events one export file holds (AUDIT_EXPORT_MAX_EVENTS_PER_FILE). Empty uses the API default of 100000."
             }
           }
         }

charts/hub/values.yaml

@@ -50,7 +50,6 @@ mongodb:
   # hub-pipeline-monitor, ...) instead of the one in uri, so the Atlas query
   # profiler and currentOp show which service issued an operation.
   appNamePerService: true
-
   # Self-hosted mongodb, you could also use the uri property above.
   host: mongodb.mongodb
   adminDatabase: admin
@@ -73,7 +72,6 @@ mongodb:
     caFileName: ""
     mountPath: "/etc/mongodb/tls"
     insecureSkipVerify: false
-
 # Classifications supported by Hub filters, alerts, markers, and other
 # downstream processing. Classifier output keys must match these values.
 # Disable this to use the legacy classification catalog stored in MongoDB.
@@ -115,7 +113,6 @@ deviceTags:
     #  - key: anpr-pilot
     #    label: ANPR pilot
     #    description: Runs licence plate recognition on this camera
-
 # Configuration packaging only: consuming contracts requires a compatible
 # API/frontend. Enabling this mounts files but does not register stage workers.
 workflowConfiguration:
@@ -339,8 +336,27 @@ kerberoshub:
   api:
     repository: ghcr.io/uug-ai/hub-api
     pullPolicy: IfNotPresent
-    tag: "v1.9.51"
+    tag: "v1.11.31"
     replicas: 2
+    podSecurityContext:
+      fsGroup: 65532
+    securityContext:
+      runAsNonRoot: true
+      runAsUser: 65532
+      runAsGroup: 65532
+      readOnlyRootFilesystem: true
+      allowPrivilegeEscalation: false
+      capabilities:
+        drop: ["ALL"]
+      seccompProfile:
+        type: RuntimeDefault
+    # Keep writable scratch space for attachment uploads and multipart parsing.
+    volumes:
+      - name: hub-api-tmp
+        emptyDir: {}
+    volumeMounts:
+      - name: hub-api-tmp
+        mountPath: /tmp
     audit:
       enabled: true
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
@@ -375,6 +391,7 @@ kerberoshub:
         memory: 100Mi
         cpu: 250m
     # E-mail templates
+    # Append custom mounts/volumes to the lists above; retain hub-api-tmp.
     #volumeMounts:
     #  - name: custom-email-templates
     #    mountPath: /mail
@@ -419,6 +436,11 @@ kerberoshub:
     # beyond this cap, the value is clamped down to `now + casesMaxRetentionDays`.
     # Set to "0" (or a negative value) to disable the cap (no maximum).
     casesMaxRetentionDays: "0"
+    # Most audit events one export file holds (AUDIT_EXPORT_MAX_EVENTS_PER_FILE).
+    # It does not refuse larger exports: the Audit page downloads them as
+    # several files of at most this size. An integer from 1 to 1000000; leave
+    # empty to use the API default of 100000.
+    auditExportMaxEventsPerFile: ""
     ## Certificates
     tls:
       - hosts:
@@ -477,11 +499,28 @@ kerberoshub:
   frontend:
     repository: ghcr.io/uug-ai/hub-frontend
     pullPolicy: IfNotPresent
-    tag: "v1.13.9"
+    tag: "v1.19.28"
     replicas: 2
-    # Hardened frontend images listen on 8080; retain 80 for existing images.
-    containerPort: 80
-    securityContext: {}
+    containerPort: 8080
+    securityContext:
+      runAsNonRoot: true
+      runAsUser: 65532
+      runAsGroup: 65532
+      readOnlyRootFilesystem: true
+      allowPrivilegeEscalation: false
+      capabilities:
+        drop: ["ALL"]
+      seccompProfile:
+        type: RuntimeDefault
+    # Keep these entries when adding custom branding/translation mounts.
+    volumes:
+      - name: frontend-tmp
+        emptyDir:
+          medium: Memory
+          sizeLimit: 16Mi
+    volumeMounts:
+      - name: frontend-tmp
+        mountPath: /tmp
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     # Set to false to skip rendering the hub-frontend Service (e.g. when an
     # external/shared Service is managed elsewhere).
@@ -771,7 +810,7 @@ kerberoshub:
     podSecurityContext: {}
     repository: ghcr.io/uug-ai/hub-cleanup
     pullPolicy: IfNotPresent
-    tag: "v1.4.19"
+    tag: "v1.4.26"
     replicas: 1 # Number of pods for the service.
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     # Optional extra volumes / volumeMounts for this deployment (empty = none).
@@ -825,7 +864,7 @@ kerberoshub:
   audit:
     repository: ghcr.io/uug-ai/hub-audit
     pullPolicy: IfNotPresent
-    tag: "v1.0.1"
+    tag: "v1.0.10"
     replicas: 1
     dispatchInterval: "5s"
     leaseDuration: "2m"
@@ -871,27 +910,31 @@ kerberoshub:
     #     actions: []
     #     categories: []
     #     outcomes: [] # success | failure
-    # S3 Object Lock destination (requires a hub-audit release with S3 support).
-    # The bucket needs versioning, Object Lock and a COMPLIANCE default
-    # retention; hub-audit refuses to deliver otherwise. Each batch is written
-    # as <prefix>/<id>/yyyy/mm/dd/<firstId>-<lastId>.ndjson.gz.
+    # S3 Object Lock destination (hub-audit v1.0.9 or later). The bucket needs
+    # versioning, Object Lock and a COMPLIANCE default retention; hub-audit
+    # refuses to deliver otherwise. Each batch is written as
+    # <prefix>/<id>/yyyy/mm/dd/<firstId>-<lastId>.ndjson.gz. Per-provider
+    # region/endpoint values (AWS, Wasabi, B2, Ceph RGW, MinIO, Garage):
+    # https://learn.uug.ai/docs/hub/audit-events/#s3-providers
     # - id: immutable-archive
     #   type: s3
     #   enabled: true
     #   timeoutSeconds: 30
     #   batchSize: 500
+    #   maxAttempts: 100 # keep high: exhausted batches are dead-lettered, not archived
     #   startAt: beginning
     #   s3:
     #     bucket: hub-audit-archive
     #     region: eu-west-1
     #     prefix: hub
-    #     endpoint: https://minio.example.com # S3-compatible stores only
-    #     forcePathStyle: true # usually needed for MinIO
+    #     endpoint: https://s3.example.com # S3-compatible stores only; omit for AWS S3
+    #     forcePathStyle: true # usually needed for Ceph RGW, MinIO and Garage
     #     # Only for stores without Object Lock (e.g. Garage): skips the lock
-    #     # check, so archived batches are deletable. Needs hub-audit with #16.
+    #     # check, so archived batches are deletable. Needs hub-audit v1.0.10.
     #     allowUnlockedBucket: false
     #   # Omit credentialsSecret to use the pod's AWS identity (IRSA via
-    #   # kerberoshub.audit.serviceAccount, or an instance profile).
+    #   # kerberoshub.audit.serviceAccount, or an instance profile). That only
+    #   # works for AWS S3; every other store needs credentialsSecret.
     #   credentialsSecret:
     #     name: audit-immutable-archive
     #     accessKeyIdKey: access-key-id
@@ -1132,13 +1175,14 @@ kerberoshub:
     podSecurityContext: {}
     repository: ghcr.io/uug-ai/hub-monitor-device
     pullPolicy: IfNotPresent
-    tag: "v1.4.2"
+    tag: "v1.4.10"
     replicas: 1 # Number of pods for the service.
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     logLevel: "info" # possible values: trace, debug, info, warn, error
     # Seconds between device status scans. Sets how quickly a camera going
-    # offline/online is detected and, with "Group notifications" on, how long
-    # notifications are held to collect other cameras that change at the same time.
+    # offline/online is detected. With "Group notifications" on, hub-monitor-device
+    # v1.4.10+ holds a change 20 seconds (DEVICE_GROUPING_HOLD_SECONDS) to collect
+    # other cameras that change at the same time; older releases hold one scan.
     scanIntervalSeconds: 60
     resources:
       requests:
@@ -1219,7 +1263,7 @@ kerberospipeline:
     #   argocd.argoproj.io/sync-wave: "10"
     repository: ghcr.io/uug-ai/hub-pipeline-event
     pullPolicy: IfNotPresent
-    tag: "v1.3.1"
+    tag: "v1.3.9"
     replicas: 1 # Number of pods for the service.
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     # Optional extra volumes / volumeMounts for this deployment (empty = none).
@@ -1240,7 +1284,7 @@ kerberospipeline:
     # deploymentAnnotations:
     #   argocd.argoproj.io/sync-wave: "70"
     pullPolicy: IfNotPresent
-    tag: "v1.3.13"
+    tag: "v1.3.29"
     replicas: 1 # Number of pods for the service.
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     # Optional extra volumes / volumeMounts for this deployment (empty = none).
@@ -1261,7 +1305,7 @@ kerberospipeline:
     # deploymentAnnotations:
     #   argocd.argoproj.io/sync-wave: "60"
     pullPolicy: IfNotPresent
-    tag: "v1.6.26"
+    tag: "v1.7.14"
     replicas: 1 # Number of pods for the service.
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     # Optional extra volumes / volumeMounts for this deployment (empty = none).
@@ -1279,12 +1323,12 @@ kerberospipeline:
   throttler:
     securityContext: {}
     podSecurityContext: {}
-    repository: uugai/hub-pipeline-throttler
+    repository: ghcr.io/uug-ai/hub-pipeline-throttler
     deploymentAnnotations: {}
     # deploymentAnnotations:
     #   argocd.argoproj.io/sync-wave: "30"
     pullPolicy: IfNotPresent
-    tag: "v1.2.1"
+    tag: "v1.2.10"
     replicas: 1 # Number of pods for the service.
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     # Optional extra volumes / volumeMounts for this deployment (empty = none).
@@ -1307,7 +1351,7 @@ kerberospipeline:
     # deploymentAnnotations:
     #   argocd.argoproj.io/sync-wave: "20"
     pullPolicy: IfNotPresent
-    tag: "v1.3.18"
+    tag: "v1.3.46"
     replicas: 1 # Number of pods for the service.
     audit:
       enabled: true
@@ -1327,9 +1371,9 @@ kerberospipeline:
         #    persistentVolumeClaim:
         #      claimName: custom-layout-claim
   notifyTest:
-    repository: uugai/hub-pipeline-notification-test
+    repository: ghcr.io/uug-ai/hub-pipeline-notification-test
     pullPolicy: IfNotPresent
-    tag: "v1.2.2"
+    tag: "v1.2.6"
     replicas: 1 # Number of pods for the service.
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     resources:
@@ -1350,7 +1394,7 @@ kerberospipeline:
     # deploymentAnnotations:
     #   argocd.argoproj.io/sync-wave: "40"
     pullPolicy: IfNotPresent
-    tag: "v1.8.22"
+    tag: "v1.8.29"
     replicas: 1 # Number of pods for the service.
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     # Optional extra volumes / volumeMounts for this deployment (empty = none).
@@ -1388,7 +1432,7 @@ kerberospipeline:
     # deploymentAnnotations:
     #   argocd.argoproj.io/sync-wave: "50"
     pullPolicy: IfNotPresent
-    tag: "v2.0.3"
+    tag: "v2.0.9"
     replicas: 3 # Number of pods for the service.
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     # Optional extra volumes / volumeMounts for this deployment (empty = none).
@@ -1414,7 +1458,7 @@ kerberospipeline:
     # deploymentAnnotations:
     #   argocd.argoproj.io/sync-wave: "50"
     pullPolicy: IfNotPresent
-    tag: "v1.3.15"
+    tag: "v1.3.22"
     replicas: 2 # Number of pods for the service.
     prefetchCount: 0 # Max unacknowledged RabbitMQ messages per pod; 0 = service default (5).
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
@@ -1442,12 +1486,12 @@ kerberospipeline:
   counting:
     securityContext: {}
     podSecurityContext: {}
-    repository: uugai/hub-pipeline-counting
+    repository: ghcr.io/uug-ai/hub-pipeline-counting
     deploymentAnnotations: {}
     # deploymentAnnotations:
     #   argocd.argoproj.io/sync-wave: "50"
     pullPolicy: IfNotPresent
-    tag: "v2.0.0"
+    tag: "v2.0.6"
     replicas: 1 # Number of pods for the service.
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     # Optional extra volumes / volumeMounts for this deployment (empty = none).
@@ -1469,7 +1513,7 @@ kerberospipeline:
     #   argocd.argoproj.io/sync-wave: "50"
     repository: ghcr.io/uug-ai/hub-pipeline-sprite
     pullPolicy: IfNotPresent
-    tag: "v1.1.21"
+    tag: "v1.1.27"
     replicas: 5 # Number of pods for the service.
     prefetchCount: 0 # Max unacknowledged RabbitMQ messages per pod; 0 = service default (5).
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
@@ -1496,7 +1540,7 @@ kerberospipeline:
   export:
     repository: ghcr.io/uug-ai/hub-pipeline-export
     pullPolicy: IfNotPresent
-    tag: "v1.2.10"
+    tag: "v1.2.20"
     replicas: 2 # Number of pods for the service.
     topologySpreadConstraints: [] # Optional pod topology spread constraints (empty = none).
     logLevel: "info" # possible values: trace, debug, info, warn, error

scripts/check-frontend-runtime.sh

@@ -0,0 +1,102 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+CHART_DIR="${1:-charts/hub}"
+values=()
+if [[ $# -ge 2 ]]; then values=(--values "$2"); fi
+
+resource() {
+  local kind="$1" name="$2"
+  awk -v kind="$kind" -v name="$name" '
+    function emit() { if (matchedKind && matchedName) printf "%s", block }
+    /^---$/ { emit(); block=""; matchedKind=0; matchedName=0 }
+    { block=block $0 "\n" }
+    $0 == "kind: " kind { matchedKind=1 }
+    $0 == "  name: " name { matchedName=1 }
+    END { emit() }
+  '
+}
+
+assert_contains() {
+  if ! grep -Eq -- "$2" <<< "$1"; then
+    printf 'Missing expected frontend setting: %s\n' "$2" >&2
+    exit 1
+  fi
+}
+
+for mode in all ui; do
+  rendered="$(helm template hub "$CHART_DIR" "${values[@]}" --set "mode=$mode" \
+    --set kerberoshub.frontend.demoEnabled=true)"
+  for name in hub-frontend hub-frontend-demo; do
+    deployment="$(resource Deployment "$name" <<< "$rendered")"
+    assert_contains "$deployment" 'image: "ghcr.io/uug-ai/hub-frontend:v1.19.28"$'
+    assert_contains "$deployment" 'containerPort: 8080$'
+    [[ "$(grep -c 'containerPort:' <<< "$deployment")" == 1 ]]
+    for setting in 'name: http' 'runAsNonRoot: true' 'runAsUser: 65532' \
+      'runAsGroup: 65532' 'readOnlyRootFilesystem: true' \
+      'allowPrivilegeEscalation: false' 'type: RuntimeDefault' 'drop:' '- ALL' \
+      'medium: Memory' 'sizeLimit: 16Mi' 'mountPath: /tmp' 'name: frontend-tmp'; do
+      assert_contains "$deployment" "^[[:space:]]*(- )?$setting$"
+    done
+    service="$(resource Service "$name-svc" <<< "$rendered")"
+    assert_contains "$service" 'port: 80$'
+    assert_contains "$service" 'targetPort: http$'
+    echo "OK ($mode): $name hardened image, security, scratch and named routing"
+  done
+  admin="$(resource Deployment admin <<< "$rendered")"
+  assert_contains "$admin" 'containerPort: 80$'
+
+  legacy="$(helm template hub "$CHART_DIR" "${values[@]}" --set "mode=$mode" \
+    --set kerberoshub.frontend.demoEnabled=true \
+    --set kerberoshub.frontend.tag=v1.13.9 \
+    --set kerberoshub.frontend.containerPort=80 \
+    --set kerberoshub.frontend.securityContext=null \
+    --set-json 'kerberoshub.frontend.volumes=[]' \
+    --set-json 'kerberoshub.frontend.volumeMounts=[]')"
+  for name in hub-frontend hub-frontend-demo; do
+    deployment="$(resource Deployment "$name" <<< "$legacy")"
+    assert_contains "$deployment" 'containerPort: 80$'
+    assert_contains "$deployment" 'image: "ghcr.io/uug-ai/hub-frontend:v1.13.9"$'
+    if grep -Eq 'securityContext:|mountPath: /tmp' <<< "$deployment"; then
+      echo "Legacy frontend retains incompatible hardened runtime settings" >&2
+      exit 1
+    fi
+  done
+done
+
+if [[ "${CLUSTER_SMOKE:-false}" == true ]]; then
+  namespace=frontend-runtime-test
+  kubectl create namespace "$namespace"
+  helm template runtime "$CHART_DIR" "${values[@]}" --namespace "$namespace" \
+    --set mode=ui --set ingress="" --set kerberoshub.frontend.demoEnabled=true \
+    --show-only templates/kerberos-hub/hub-frontend.yaml \
+    --show-only templates/kerberos-hub/hub-frontend-demo.yaml | kubectl apply -f -
+  pid=""
+  cleanup() {
+    if [[ -n "$pid" ]]; then kill "$pid"; fi
+    if [[ "$1" != 0 ]]; then
+      kubectl -n "$namespace" get pods
+      kubectl -n "$namespace" describe pods
+      kubectl -n "$namespace" logs -l app=hub-frontend --all-containers=true
+      kubectl -n "$namespace" logs -l app=hub-frontend-demo --all-containers=true
+    fi
+  }
+  trap 'cleanup "$?"' EXIT
+  for name in hub-frontend hub-frontend-demo; do
+    kubectl -n "$namespace" rollout status "deployment/$name" --timeout=180s
+    kubectl -n "$namespace" port-forward "service/$name-svc" 18080:80 &
+    pid=$!
+    for attempt in {1..30}; do
+      if curl --fail --silent http://127.0.0.1:18080/healthz > /dev/null; then break; fi
+      sleep 1
+    done
+    [[ "$(curl --fail --silent http://127.0.0.1:18080/healthz)" == ok ]]
+    curl --fail --silent http://127.0.0.1:18080/assets/env.js | grep 'window' > /dev/null
+    kill "$pid"
+    wait "$pid" || [[ "$?" == 143 ]]
+    pid=""
+    echo "OK: $name serves health and generated runtime configuration through Service port 80"
+  done
+fi
+
+echo "Frontend runtime checks passed."

scripts/check-hardened-workers.sh

@@ -60,10 +60,19 @@ assert_line() {
   fi
 }
 
+without_frontends() {
+  awk '
+    /^---$/ { skip=0 }
+    /^# Source: hub\/templates\/kerberos-hub\/(hub-frontend(-demo)?|hub-api)\.yaml$/ { skip=1 }
+    !skip { print }
+  '
+}
+
 for mode in all pipeline ui; do
   defaults="$(helm template hub "$CHART_DIR" --set "mode=$mode")"
-  if grep -q 'securityContext:' <<< "$defaults"; then
-    echo "Default $mode mode unexpectedly enables security settings" >&2
+  worker_defaults="$(without_frontends <<< "$defaults")"
+  if grep -q 'securityContext:' <<< "$worker_defaults"; then
+    echo "Default $mode mode unexpectedly enables worker security settings" >&2
     exit 1
   fi
 
@@ -111,8 +120,8 @@ for mode in all pipeline ui; do
     fi
   done
   expected="${#names[@]}"
-  [[ "$(grep -c 'runAsUser: 65532' <<< "$hardened")" == "$expected" ]]
-  [[ "$(grep -c 'fsGroup: 65532' <<< "$hardened")" == "$expected" ]]
+  [[ "$(without_frontends <<< "$hardened" | grep -c 'runAsUser: 65532')" == "$expected" ]]
+  [[ "$(without_frontends <<< "$hardened" | grep -c 'fsGroup: 65532')" == "$expected" ]]
 
   # Image selection, Services, env and TLS mounts must not change.
   for pattern in '^[ ]+image:' '^[ ]+(port|targetPort|containerPort):' 'mongodb-tls'; do
@@ -179,7 +188,7 @@ for mode in all ui pipeline; do
   [[ "$(resource Service oauth2-proxy-admin <<< "$legacy")" == "$(resource Service oauth2-proxy-admin <<< "$configured")" ]]
   [[ "$(grep 'image:' <<< "$legacy")" == "$(grep 'image:' <<< "$configured")" ]]
   # No other workload inherits Admin's container port or security settings.
-  [[ "$(grep -c 'runAsUser: 65532' <<< "$configured")" == 1 ]]
+  [[ "$(without_frontends <<< "$configured" | grep -c 'runAsUser: 65532')" == 1 ]]
   echo "OK ($mode): Admin opt-in port/security/tmpfs; legacy image, Service port, ingress and OAuth2-proxy preserved"
 done
 echo "All hardened Admin checks passed."

scripts/check-http-port-migration.sh

@@ -0,0 +1,100 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
+REAL_KUBECTL="$(command -v kubectl)"
+export REAL_KUBECTL
+
+# Substitute only transport: exercise the migration against Kubernetes' real patch engine.
+kubectl() {
+  if [[ "$*" == *" get deployment "* ]]; then
+    printf '%s\n' "$LIVE"
+  else
+    local patch="" previous="" argument
+    for argument in "$@"; do
+      if [[ "$previous" == -p ]]; then patch="$argument"; fi
+      previous="$argument"
+    done
+    if [[ "${EXPECT_APPLY:-false}" == true ]]; then
+      [[ "$*" != *"--dry-run=server"* ]] || return 1
+    else
+      [[ "$*" == *"--dry-run=server"* ]] || return 1
+    fi
+    local current="$LIVE"
+    if [[ "${SIMULATE_RACE:-false}" == true ]]; then
+      current="$(jq '.metadata.resourceVersion = "2"' <<< "$LIVE")"
+    fi
+    "$REAL_KUBECTL" patch --local -f <(printf '%s\n' "$current") \
+      --type=json -p "$patch" -o json
+  fi
+}
+export -f kubectl
+
+for deployment in admin hub-frontend hub-frontend-demo; do
+  for old in 80 8080; do
+    target=8080
+    if [[ "$old" == 8080 ]]; then target=80; fi
+    LIVE="$(jq -nc --arg name "$deployment" --argjson port "$old" '{
+      apiVersion: "apps/v1", kind: "Deployment",
+      metadata: {name: $name, resourceVersion: "1"},
+      spec: {selector: {matchLabels: {app: $name}}, template: {
+        metadata: {labels: {app: $name}}, spec: {containers: [
+          {name: "sidecar", image: "example/sidecar:unchanged"},
+          {name: $name, image: "example/frontend:unchanged", ports: [
+            {name: "http", containerPort: $port},
+            {name: "metrics", containerPort: 2112}
+          ]}
+        ]}
+      }}
+    }')"
+    export LIVE
+    merge="$(jq -nc --arg name "$deployment" --argjson port "$target" \
+      '{spec:{template:{spec:{containers:[{name:$name,ports:[{name:"http",containerPort:$port}]}]}}}}')"
+    merged="$("$REAL_KUBECTL" patch --local -f <(printf '%s\n' "$LIVE") \
+      --type=strategic -p "$merge" -o json)"
+    jq -e '[.spec.template.spec.containers[] | .ports[]? |
+      select(.name == "http")] | length == 2' <<< "$merged" > /dev/null
+
+    fixed="$(bash "$ROOT/scripts/migrate-http-port.sh" test "$deployment" "$target")"
+    jq -e --argjson port "$target" '
+      [.spec.template.spec.containers[] | .ports[]? | select(.name == "http")] |
+      length == 1 and .[0].containerPort == $port
+    ' <<< "$fixed" > /dev/null
+    [[ "$(jq -Sc '.spec.template.spec.containers[0]' <<< "$fixed")" == \
+       "$(jq -Sc '.spec.template.spec.containers[0]' <<< "$LIVE")" ]]
+    [[ "$(jq -Sc '.spec.template.spec.containers[1] | del(.ports)' <<< "$fixed")" == \
+       "$(jq -Sc '.spec.template.spec.containers[1] | del(.ports)' <<< "$LIVE")" ]]
+    jq -e '.spec.template.spec.containers[1].ports |
+      any(.name == "metrics" and .containerPort == 2112)' <<< "$fixed" > /dev/null
+
+    LIVE="$fixed"
+    again="$(EXPECT_APPLY=true bash "$ROOT/scripts/migrate-http-port.sh" test "$deployment" "$target" --apply)"
+    [[ "$(jq -Sc . <<< "$again")" == "$(jq -Sc . <<< "$fixed")" ]]
+    LIVE="$merged"
+    repaired="$(EXPECT_APPLY=true bash "$ROOT/scripts/migrate-http-port.sh" test "$deployment" "$target" --apply)"
+    jq -e '[.spec.template.spec.containers[] | .ports[]? |
+      select(.name == "http")] | length == 1' <<< "$repaired" > /dev/null
+
+    if [[ "${SERVER_VALIDATION:-false}" == true ]]; then
+      if "$REAL_KUBECTL" create --dry-run=server \
+        -f <(jq 'del(.metadata.resourceVersion)' <<< "$merged") > /dev/null 2>&1; then
+        echo "Expected API rejection of duplicate http ports" >&2
+        exit 1
+      fi
+      "$REAL_KUBECTL" create --dry-run=server \
+        -f <(jq 'del(.metadata.resourceVersion)' <<< "$repaired") > /dev/null
+    fi
+    echo "OK: $deployment $old -> $target; duplicate reproduced, replacement and rollback safe"
+  done
+done
+
+if SIMULATE_RACE=true bash "$ROOT/scripts/migrate-http-port.sh" test hub-frontend-demo 8080 > /dev/null 2>&1; then
+  echo "Expected rejection of concurrent Deployment changes" >&2
+  exit 1
+fi
+LIVE="$(jq '.spec.template.spec.containers[1].ports = [{name:"http",containerPort:9000}]' <<< "$LIVE")"
+if bash "$ROOT/scripts/migrate-http-port.sh" test hub-frontend-demo 8080 > /dev/null 2>&1; then
+  echo "Expected rejection of an unknown HTTP port" >&2
+  exit 1
+fi
+echo "HTTP port migration checks passed."

scripts/check-hub-api-runtime.sh

@@ -0,0 +1,86 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+CHART_DIR="${1:-charts/hub}"
+
+resource() {
+  local kind="$1" name="$2"
+  awk -v kind="$kind" -v name="$name" '
+    function emit() { if (matchedKind && matchedName) printf "%s", block }
+    /^---$/ { emit(); block=""; matchedKind=0; matchedName=0 }
+    { block=block $0 "\n" }
+    $0 == "kind: " kind { matchedKind=1 }
+    $0 == "  name: " name { matchedName=1 }
+    END { emit() }
+  '
+}
+
+assert_line() {
+  if ! grep -Eq -- "$2" <<< "$1"; then
+    printf 'Missing expected Hub API setting: %s\n' "$2" >&2
+    exit 1
+  fi
+}
+
+fixtures=(
+  --set workflowConfiguration.enabled=true
+  --set mongodb.tls.enabled=true
+  --set mongodb.tls.existingSecret=mongodb-ca
+  --set kerberoshub.api.serverTLS.enabled=true
+  --set kerberoshub.api.serverTLS.secretName=api-tls
+)
+
+for mode in all ui pipeline; do
+  defaults="$(helm template hub "$CHART_DIR" --set "mode=$mode" "${fixtures[@]}")"
+  legacy="$(helm template hub "$CHART_DIR" --set "mode=$mode" "${fixtures[@]}" \
+    --set kerberoshub.api.securityContext=null \
+    --set kerberoshub.api.podSecurityContext=null)"
+  configured="$(helm template hub "$CHART_DIR" --set "mode=$mode" "${fixtures[@]}" \
+    --set kerberoshub.api.securityContext.runAsUser=1000 \
+    --set kerberoshub.api.securityContext.runAsGroup=1000 \
+    --set kerberoshub.api.podSecurityContext.fsGroup=1000)"
+  helm lint "$CHART_DIR" --quiet --set "mode=$mode" "${fixtures[@]}"
+
+  if [[ "$mode" == pipeline ]]; then
+    [[ -z "$(resource Deployment hub-api <<< "$defaults")" ]]
+    [[ "$defaults" == "$legacy" && "$defaults" == "$configured" ]]
+    echo "OK (pipeline): no Hub API resources or unrelated changes"
+    continue
+  fi
+
+  api="$(resource Deployment hub-api <<< "$defaults")"
+  for pattern in \
+    '^[ ]{6}securityContext:$' '^[ ]{8}fsGroup: 65532$' \
+    '^[ ]{10}securityContext:$' '^[ ]{12}runAsUser: 65532$' \
+    '^[ ]{12}runAsGroup: 65532$' '^[ ]{12}runAsNonRoot: true$' \
+    '^[ ]{12}readOnlyRootFilesystem: true$' \
+    '^[ ]{12}allowPrivilegeEscalation: false$' '^[ ]{12}capabilities:$' \
+    '^[ ]{14}drop:$' '^[ ]{14}- ALL$' \
+    '^[ ]{12}seccompProfile:$' '^[ ]{14}type: RuntimeDefault$' \
+    '^[ ]+(- )?name: hub-api-tmp$' '^[ ]+(- )?emptyDir: \{\}$' \
+    '^[ ]+(- )?mountPath: /tmp$' '^[ ]+secretName: mongodb-ca$' \
+    '^[ ]+secretName: api-tls$' '^[ ]+mountPath: /etc/mongodb/tls$' \
+    '^[ ]+mountPath: /etc/hub-api/tls$' \
+    '^[ ]+mountPath: /etc/kerberos/configuration$' \
+    '^[ ]+mountPath: /etc/kerberos/classifications$'; do
+    assert_line "$api" "$pattern"
+  done
+  [[ "$(grep -c 'securityContext:' <<< "$api")" == 2 ]]
+  old_api="$(resource Deployment hub-api <<< "$legacy")"
+  if grep -q 'securityContext:' <<< "$old_api"; then
+    echo "Legacy Hub API opt-out still renders a security context" >&2
+    exit 1
+  fi
+  custom_api="$(resource Deployment hub-api <<< "$configured")"
+  for field in runAsUser runAsGroup fsGroup; do
+    assert_line "$custom_api" "^[ ]+$field: 1000$"
+  done
+
+  # Only security settings change; preserve env, mounts, images and routing.
+  for pattern in '^[ ]+image:' '^[ ]+(port|targetPort|containerPort):' \
+    '^[ ]+(- )?(name|mountPath|secretName|value):'; do
+    [[ "$(grep -E "$pattern" <<< "$defaults")" == "$(grep -E "$pattern" <<< "$legacy")" ]]
+    [[ "$(grep -E "$pattern" <<< "$defaults")" == "$(grep -E "$pattern" <<< "$configured")" ]]
+  done
+  echo "OK ($mode): API security defaults, overrides, opt-out, scratch and TLS/config mounts"
+done

scripts/migrate-http-port.sh

@@ -0,0 +1,54 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+if [[ $# -lt 3 || $# -gt 4 ]]; then
+  echo "Usage: bash scripts/migrate-http-port.sh NAMESPACE DEPLOYMENT PORT [--apply]" >&2
+  exit 1
+fi
+namespace="$1" deployment="$2" port="$3"
+case "$deployment" in
+  admin|hub-frontend|hub-frontend-demo) ;;
+  *) echo "Unsupported Deployment: $deployment" >&2; exit 1 ;;
+esac
+case "$port" in
+  80|8080) ;;
+  *) echo "Expected HTTP port 80 or 8080, got: $port" >&2; exit 1 ;;
+esac
+dry_run=(--dry-run=server)
+if [[ $# -eq 4 ]]; then
+  if [[ "$4" != --apply ]]; then
+    echo "Unknown option: $4" >&2
+    exit 1
+  fi
+  dry_run=()
+fi
+
+live="$(kubectl -n "$namespace" get deployment "$deployment" -o json)"
+patch="$(jq -ce --arg container "$deployment" --argjson port "$port" '
+  . as $deployment |
+  [.spec.template.spec.containers | to_entries[] |
+    select(.value.name == $container)] |
+  if length != 1 then error("Expected exactly one matching container") else .[0] end |
+  .key as $index | .value.ports as $ports |
+  if ($ports | type) != "array" then error("Container ports must be an array") else . end |
+  [$ports[] | select(.name == "http")] as $http |
+  if ($http | length) == 0 then error("No named http port to migrate") else . end |
+  if any($http[]; (.containerPort != 80 and .containerPort != 8080) or
+    ((.protocol // "TCP") != "TCP") or has("hostPort") or has("hostIP"))
+  then error("Unsupported http port configuration; migrate manually") else . end |
+  if any($ports[]; .name != "http" and .containerPort == $port and
+    ((.protocol // "TCP") == "TCP"))
+  then error("Target port is already used by another named port") else . end |
+  ([$ports[] | select(.name != "http")] +
+    [($http[0] + {containerPort: $port})]) as $replacement |
+  [
+    {op: "test", path: "/metadata/resourceVersion", value: $deployment.metadata.resourceVersion},
+    {op: "test", path: ("/spec/template/spec/containers/" + ($index | tostring) + "/name"),
+      value: $container},
+    {op: "replace", path: ("/spec/template/spec/containers/" + ($index | tostring) + "/ports"),
+      value: $replacement}
+  ]
+' <<< "$live")"
+
+kubectl -n "$namespace" patch deployment "$deployment" --type=json \
+  -p "$patch" "${dry_run[@]}" -o yaml