test: e2e tests for status.url, k8s proxy access, and SSH edge URL (#73)

* test: add e2e tests for status.url, k8s proxy, and SSH edge URL

- Add EdgeURLSet: verifies edge.status.url populated for kubernetes-type
- Add K8sProxyAccess: verifies kubectl works against status.url
- Add SSHEdgeURLSet: verifies status.url set for server-type edges
- Fix AgentEdgeJoin: use status.url for proxy access assertion
- Add GetEdgeURL + KubectlWithURL framework helpers
- Fix mount_reconciler: set SSH URL for server-type edges instead of clearing
- Sync CRD YAML: update URL field description to match current types

Co-authored-by: Mangirdas Judeikis <mangirdas@judeikis.lt>

* test: fix e2e tests - verified passing locally

Fixes all failures in the SSH and standalone e2e suites.

## Root causes fixed

### 1. Port conflict: KEDGE_HUB_API_SERVER_PORT (framework/cluster.go)
Port 6443 was occupied by kcp on this host. Add
KEDGE_HUB_API_SERVER_PORT env-var support so  can
be told to use a different API server port (e.g. 6444).
All SetupClusters* variants now forward the flag.

### 2. SSH WebSocket URL double-concatenation (pkg/cli/cmd/ssh.go)
buildSSHWebSocketURL was prepending the hub base path on top of the
full status.URL, producing nonsense URLs like:
  wss://hub:8443/services/edges-proxyhttps://hub:8443/.../ssh/ssh
Now it simply parses status.URL (already the complete URL) and
converts the scheme to wss.

### 3. Cluster name mismatch for static-token agents (tunneler.go / agent.go)
extractClusterNameFromToken returned "default" for non-JWT static
tokens. The hub's mount_reconciler embeds the real kcp cluster name
(e.g. "1tww43gelbj45g0k") in status.URL; the agent was registering
its reverse-tunnel under the key edges/default/... causing the SSH
and k8s proxy handlers to return 502.

Fix: add SplitBaseAndCluster(hubURL) which extracts the cluster from
the kubeconfig server URL (set by static-token login as
https://hub:8443/clusters/<id>). Priority: explicit --cluster flag >
URL-embedded cluster > JWT claim > "default" fallback. Also add
--cluster flag to  for the Docker container case.

### 4. ServerProcess uses hub kubeconfig (test/e2e/framework/server_process.go)
SSH server-mode agents were started with bare --hub-url (no cluster
path). Add HubKubeconfig field: when set, use --hub-kubeconfig +
--tunnel-url so the agent gets the cluster-scoped server URL and
SplitBaseAndCluster picks up the correct cluster ID.

### 5. DialSSH hardcoded /clusters/default (framework/ssh_client.go)
Look up edge.Status.URL (which has the correct cluster) then convert
https→wss instead of constructing the URL from scratch with a
hardcoded "default" cluster path.

### 6. VirtualWorkload siteSelector → edgeSelector (cases/multisite.go)
PlacementSpec.EdgeSelector is the actual field name; siteSelector
caused strict-decoding 400 errors for LabelBasedScheduling,
WorkloadIsolation, SiteFailoverIsolation, SiteReconnect tests.

Co-authored-by: Mangirdas Judeikis <mangirdas@judeikis.lt>

* test: add functional e2e tests for workload deployment and k8s proxy

- WorkloadDeployment: VirtualWorkload → Placement → resource on edge cluster
- K8sProxyWrite: kubectl apply via edges-proxy creates resource on edge
- K8sProxyExec: kubectl exec via edges-proxy (tests upgrade/hijack path)

Also fix a bug in edgesHandleK8sUpgrade: the request URL path was not
being rewritten to the /k8s/... form before forwarding to the agent,
causing the agent's mux to return 404 for exec/port-forward requests.
Now we call extractEdgeK8sPath before r.Write(deviceConn).

All tested locally against real kind clusters.

Co-authored-by: Mangirdas Judeikis <mangirdas@judeikis.lt>

* fix: rename TestOIDCUserCanListSites → TestOIDCUserCanListEdges, use kedge edge list

kedge site list was removed; the OIDC test must use kedge edge list.

Co-authored-by: Mangirdas Judeikis <mangirdas@judeikis.lt>

* fix: rename --site-name → --edge-name, PlacementSpec.SiteName → EdgeName

Addresses PR #73 review comment. Sites are gone; use edge terminology:
- CLI flag --site-name → --edge-name (deprecated alias kept)
- PlacementSpec.SiteName json:siteName → EdgeName json:edgeName
- All scheduler/aggregator/workload reconciler references updated
- Test framework updated

Co-authored-by: Mangirdas Judeikis <mangirdas@judeikis.lt>

* fix(e2e): use edgeName json field in WaitForPlacement kubectl query

.spec.siteName was renamed to .spec.edgeName when PlacementSpec was
migrated from Site to Edge CRD. Update both WaitForPlacement and
WaitForNoPlacement custom-columns queries to match.

Co-authored-by: Mangirdas Judeikis <mangirdas@judeikis.lt>

* refactor: remove all stale site/siteName references

- Drop --site-name and --mode deprecated CLI flags (no backward compat)
- Remove AgentModeSite, AgentMode, AgentModeServer compat constants/types
- Rename kedge.faros.sh/site label -> kedge.faros.sh/edge in scheduler
  and agent workload reconciler
- Rename siteName fields/params -> edgeName throughout agent and test packages
- Fix kedge get placements: column SITE->EDGE, spec.siteName->spec.edgeName
- Fix WaitForPlacement/WaitForNoPlacement: custom-columns SITE->EDGE
- Fix Helm dev plugin examples: siteName->edgeName, site-kubeconfig->edge-kubeconfig
- Update 'kedge site list/get' examples -> 'kedge edge list/get'

Co-authored-by: Mangirdas Judeikis <mangirdas@judeikis.lt>

* refactor: replace remaining site references with edge throughout codebase

Update all remaining "site" references from the old Site CRD concept:
- Rename test data edge names (site-1 → edge-1, etc.)
- Rename functions SiteFailoverIsolation → EdgeFailoverIsolation, etc.
- Rename constants msSite1 → msEdge1, etc.
- Fix user-visible strings in dev plugin (site kubeconfig → edge kubeconfig)
- Update all comments and package docs
- Rename test functions NoMatchingSites → NoMatchingEdges, etc.

Co-authored-by: Mangirdas Judeikis <mangirdas@judeikis.lt>

* chore: regenerate CRD manifests after virtualworkload doc update

Update generated CRDs and APIResourceSchema to reflect description
changes: 'across sites' -> 'across edges'.

Co-authored-by: Mangirdas Judeikis <mangirdas@judeikis.lt>

---------

Co-authored-by: OpenClaw Bot <bot@openclaw.ai>
Co-authored-by: Mangirdas Judeikis <mangirdas@judeikis.lt>

Author: 3 people

apis/kedge/v1alpha1/types_placement.go

@@ -23,7 +23,7 @@ import (
 
 // +genclient
 // +kubebuilder:subresource:status
-// +kubebuilder:printcolumn:name="Edge",type="string",JSONPath=".spec.siteName"
+// +kubebuilder:printcolumn:name="Edge",type="string",JSONPath=".spec.edgeName"
 // +kubebuilder:printcolumn:name="Phase",type="string",JSONPath=".status.phase"
 // +kubebuilder:printcolumn:name="Ready",type="integer",JSONPath=".status.readyReplicas"
 // +kubebuilder:printcolumn:name="Age",type="date",JSONPath=".metadata.creationTimestamp"
@@ -49,7 +49,7 @@ type PlacementList struct {
 // PlacementObjSpec defines the desired state of a Placement.
 type PlacementObjSpec struct {
 	WorkloadRef corev1.ObjectReference `json:"workloadRef"`
-	SiteName    string                 `json:"siteName"`
+	EdgeName    string                 `json:"edgeName"`
 	Replicas    *int32                 `json:"replicas,omitempty"`
 }
 

apis/kedge/v1alpha1/types_virtualworkload.go

@@ -31,7 +31,7 @@ const (
 	VirtualWorkloadPhaseUnknown VirtualWorkloadPhase = "Unknown"
 )
 
-// PlacementStrategy defines how workloads are placed across sites.
+// PlacementStrategy defines how workloads are placed across edges.
 type PlacementStrategy string
 
 const (
@@ -48,7 +48,7 @@ const (
 // +kubebuilder:printcolumn:name="Age",type="date",JSONPath=".metadata.creationTimestamp"
 // +k8s:deepcopy-gen:interfaces=k8s.io/apimachinery/pkg/runtime.Object
 
-// VirtualWorkload describes a workload to be deployed across sites.
+// VirtualWorkload describes a workload to be deployed across edges.
 type VirtualWorkload struct {
 	metav1.TypeMeta   `json:",inline"`
 	metav1.ObjectMeta `json:"metadata,omitempty"`

cmd/kedge-agent/main.go

@@ -35,7 +35,7 @@ func main() {
 
 	cmd := &cobra.Command{
 		Use:   "kedge-agent",
-		Short: "Kedge agent - connects a site to the hub via reverse tunnel",
+		Short: "Kedge agent - connects an edge to the hub via reverse tunnel",
 		RunE: func(cmd *cobra.Command, args []string) error {
 			ctx, cancel := signal.NotifyContext(context.Background(), os.Interrupt, syscall.SIGTERM)
 			defer cancel()
@@ -54,18 +54,12 @@ func main() {
 	cmd.Flags().StringVar(&opts.HubContext, "hub-context", "", "Kubeconfig context for hub cluster")
 	cmd.Flags().StringVar(&opts.TunnelURL, "tunnel-url", "", "Hub tunnel URL (defaults to hub URL)")
 	cmd.Flags().StringVar(&opts.Token, "token", "", "Bootstrap token")
-	cmd.Flags().StringVar(&opts.SiteName, "site-name", "", "Name of this site")
+	cmd.Flags().StringVar(&opts.EdgeName, "edge-name", "", "Name of this edge")
 	cmd.Flags().StringVar(&opts.Kubeconfig, "kubeconfig", "", "Path to target cluster kubeconfig")
 	cmd.Flags().StringVar(&opts.Context, "context", "", "Kubeconfig context to use")
-	cmd.Flags().StringToStringVar(&opts.Labels, "labels", nil, "Labels for this site (key=value pairs)")
+	cmd.Flags().StringToStringVar(&opts.Labels, "labels", nil, "Labels for this edge (key=value pairs)")
 	cmd.Flags().StringVar((*string)(&opts.Type), "type", string(agent.AgentTypeKubernetes),
 		"Edge type: 'kubernetes' (k8s cluster) or 'server' (bare-metal/systemd host)")
-	// --mode is a deprecated alias for --type; kept for backward compatibility.
-	cmd.Flags().StringVar((*string)(&opts.Mode), "mode", "", //nolint:staticcheck
-		"Deprecated: use --type. Agent mode: 'site' (→ kubernetes) or 'server'")
-	if err := cmd.Flags().MarkDeprecated("mode", "use --type instead (kubernetes|server)"); err != nil {
-		klog.Warning("failed to mark --mode as deprecated: ", err)
-	}
 	cmd.Flags().BoolVar(&opts.InsecureSkipTLSVerify, "insecure-skip-tls-verify", false, "Skip TLS certificate verification for hub connection (dev/test only)")
 	cmd.Flags().IntVar(&opts.SSHProxyPort, "ssh-proxy-port", 22, "Local SSH daemon port to proxy connections to")
 	cmd.Flags().StringVar(&opts.SSHUser, "ssh-user", "", "SSH username for server-type edges (default: current user)")

config/crds/kedge.faros.sh_edges.yaml

@@ -117,8 +117,10 @@ spec:
             properties:
               URL:
                 description: |-
-                  WorkspaceURL is the virtual workspace URL for this edge.
-                  Only set for type=kubernetes edges.
+                  URL is the proxy URL path for accessing this edge via the hub.
+                  Format: /clusters/{cluster}/apis/kedge.faros.sh/v1alpha1/edges/{name}
+                  For kubernetes edges, append /k8s/ for K8s API proxy.
+                  For server edges, append /ssh for SSH WebSocket terminal.
                 type: string
               connected:
                 description: Connected indicates whether the edge agent currently

config/crds/kedge.faros.sh_placements.yaml

@@ -15,7 +15,7 @@ spec:
   scope: Namespaced
   versions:
   - additionalPrinterColumns:
-    - jsonPath: .spec.siteName
+    - jsonPath: .spec.edgeName
       name: Edge
       type: string
     - jsonPath: .status.phase
@@ -52,11 +52,11 @@ spec:
           spec:
             description: PlacementObjSpec defines the desired state of a Placement.
             properties:
+              edgeName:
+                type: string
               replicas:
                 format: int32
                 type: integer
-              siteName:
-                type: string
               workloadRef:
                 description: ObjectReference contains enough information to let you
                   inspect or modify the referred object.
@@ -102,7 +102,7 @@ spec:
                 type: object
                 x-kubernetes-map-type: atomic
             required:
-            - siteName
+            - edgeName
             - workloadRef
             type: object
           status:

config/crds/kedge.faros.sh_virtualworkloads.yaml

@@ -32,7 +32,7 @@ spec:
     name: v1alpha1
     schema:
       openAPIV3Schema:
-        description: VirtualWorkload describes a workload to be deployed across sites.
+        description: VirtualWorkload describes a workload to be deployed across edges.
         properties:
           apiVersion:
             description: |-
@@ -119,7 +119,7 @@ spec:
                     x-kubernetes-map-type: atomic
                   strategy:
                     description: PlacementStrategy defines how workloads are placed
-                      across sites.
+                      across edges.
                     type: string
                 type: object
               replicas:

config/kcp/apiexport-kedge.faros.sh.yaml

@@ -65,12 +65,12 @@ spec:
   resources:
   - group: kedge.faros.sh
     name: edges
-    schema: v260228-0f5e1a9.edges.kedge.faros.sh
+    schema: v260228-eea5d12.edges.kedge.faros.sh
     storage:
       crd: {}
   - group: kedge.faros.sh
     name: placements
-    schema: v260227-f675e28.placements.kedge.faros.sh
+    schema: v260301-c08d8fa.placements.kedge.faros.sh
     storage:
       crd: {}
   - group: kedge.faros.sh
@@ -80,7 +80,7 @@ spec:
       crd: {}
   - group: kedge.faros.sh
     name: virtualworkloads
-    schema: v260227-f675e28.virtualworkloads.kedge.faros.sh
+    schema: v260301-1a00c73.virtualworkloads.kedge.faros.sh
     storage:
       crd: {}
 status: {}

config/kcp/apiresourceschema-edges.kedge.faros.sh.yaml

@@ -1,7 +1,7 @@
 apiVersion: apis.kcp.io/v1alpha1
 kind: APIResourceSchema
 metadata:
-  name: v260228-0f5e1a9.edges.kedge.faros.sh
+  name: v260228-eea5d12.edges.kedge.faros.sh
 spec:
   group: kedge.faros.sh
   names:
@@ -113,8 +113,10 @@ spec:
           properties:
             URL:
               description: |-
-                WorkspaceURL is the virtual workspace URL for this edge.
-                Only set for type=kubernetes edges.
+                URL is the proxy URL path for accessing this edge via the hub.
+                Format: /clusters/{cluster}/apis/kedge.faros.sh/v1alpha1/edges/{name}
+                For kubernetes edges, append /k8s/ for K8s API proxy.
+                For server edges, append /ssh for SSH WebSocket terminal.
               type: string
             connected:
               description: Connected indicates whether the edge agent currently has

config/kcp/apiresourceschema-placements.kedge.faros.sh.yaml

@@ -1,7 +1,7 @@
 apiVersion: apis.kcp.io/v1alpha1
 kind: APIResourceSchema
 metadata:
-  name: v260227-f675e28.placements.kedge.faros.sh
+  name: v260301-c08d8fa.placements.kedge.faros.sh
 spec:
   group: kedge.faros.sh
   names:
@@ -12,7 +12,7 @@ spec:
   scope: Namespaced
   versions:
   - additionalPrinterColumns:
-    - jsonPath: .spec.siteName
+    - jsonPath: .spec.edgeName
       name: Edge
       type: string
     - jsonPath: .status.phase
@@ -48,11 +48,11 @@ spec:
         spec:
           description: PlacementObjSpec defines the desired state of a Placement.
           properties:
+            edgeName:
+              type: string
             replicas:
               format: int32
               type: integer
-            siteName:
-              type: string
             workloadRef:
               description: ObjectReference contains enough information to let you
                 inspect or modify the referred object.
@@ -98,7 +98,7 @@ spec:
               type: object
               x-kubernetes-map-type: atomic
           required:
-          - siteName
+          - edgeName
           - workloadRef
           type: object
         status:

config/kcp/apiresourceschema-virtualworkloads.kedge.faros.sh.yaml

@@ -1,7 +1,7 @@
 apiVersion: apis.kcp.io/v1alpha1
 kind: APIResourceSchema
 metadata:
-  name: v260227-f675e28.virtualworkloads.kedge.faros.sh
+  name: v260301-1a00c73.virtualworkloads.kedge.faros.sh
 spec:
   group: kedge.faros.sh
   names:
@@ -28,7 +28,7 @@ spec:
       type: date
     name: v1alpha1
     schema:
-      description: VirtualWorkload describes a workload to be deployed across sites.
+      description: VirtualWorkload describes a workload to be deployed across edges.
       properties:
         apiVersion:
           description: |-
@@ -115,7 +115,7 @@ spec:
                   x-kubernetes-map-type: atomic
                 strategy:
                   description: PlacementStrategy defines how workloads are placed
-                    across sites.
+                    across edges.
                   type: string
               type: object
             replicas: