fix: allocate fresh network on start VM

.github/workflows/test.yml

@@ -34,6 +34,9 @@ jobs:
           username: ${{ secrets.DOCKERHUB_USERNAME }}
           password: ${{ secrets.DOCKERHUB_PASSWORD }}
 
+      - name: Clean cached binaries
+        run: make clean
+
       - name: Build
         run: make build
 

Makefile

@@ -1,5 +1,5 @@
 SHELL := /bin/bash
-.PHONY: oapi-generate generate-vmm-client generate-wire generate-all dev build test install-tools gen-jwt download-ch-binaries download-ch-spec ensure-ch-binaries build-caddy-binaries build-caddy ensure-caddy-binaries release-prep
+.PHONY: oapi-generate generate-vmm-client generate-wire generate-all dev build test install-tools gen-jwt download-ch-binaries download-ch-spec ensure-ch-binaries build-caddy-binaries build-caddy ensure-caddy-binaries release-prep clean
 
 # Directory where local binaries will be installed
 BIN_DIR ?= $(CURDIR)/bin
@@ -192,10 +192,8 @@ gen-jwt: $(GODOTENV)
 # Clean generated files and binaries
 clean:
 	rm -rf $(BIN_DIR)
-	rm -f lib/oapi/oapi.go
-	rm -f lib/vmm/vmm.go
-	rm -f lib/exec/exec.pb.go
-	rm -f lib/exec/exec_grpc.pb.go
+	rm -rf lib/vmm/binaries/cloud-hypervisor/
+	rm -rf lib/ingress/binaries/
 	rm -f lib/system/exec_agent/exec-agent
 
 # Prepare for release build (called by GoReleaser)

lib/instances/start.go

@@ -8,6 +8,7 @@ import (
 	"github.com/onkernel/hypeman/lib/logger"
 	"github.com/onkernel/hypeman/lib/network"
 	"go.opentelemetry.io/otel/trace"
+	"gvisor.dev/gvisor/pkg/cleanup"
 )
 
 // startInstance starts a stopped instance
@@ -52,46 +53,53 @@ func (m *manager) startInstance(
 		return nil, fmt.Errorf("get image: %w", err)
 	}
 
-	// 4. Recreate network allocation if network enabled
+	// Setup cleanup stack for automatic rollback on errors
+	cu := cleanup.Make(func() {})
+	defer cu.Clean()
+
+	// 4. Allocate fresh network if network enabled
 	var netConfig *network.NetworkConfig
 	if stored.NetworkEnabled {
-		log.DebugContext(ctx, "recreating network for start", "instance_id", id, "network", "default")
-		if err := m.networkManager.RecreateAllocation(ctx, id); err != nil {
-			log.ErrorContext(ctx, "failed to recreate network", "instance_id", id, "error", err)
-			return nil, fmt.Errorf("recreate network: %w", err)
-		}
-		// Get the network config for VM configuration
-		netAlloc, err := m.networkManager.GetAllocation(ctx, id)
+		log.DebugContext(ctx, "allocating network for start", "instance_id", id, "network", "default")
+		netConfig, err = m.networkManager.CreateAllocation(ctx, network.AllocateRequest{
+			InstanceID:   id,
+			InstanceName: stored.Name,
+		})
 		if err != nil {
-			log.ErrorContext(ctx, "failed to get network allocation", "instance_id", id, "error", err)
-			// Cleanup network on failure
-			if netAlloc != nil {
-				m.networkManager.ReleaseAllocation(ctx, netAlloc)
-			}
-			return nil, fmt.Errorf("get network allocation: %w", err)
-		}
-		netConfig = &network.NetworkConfig{
-			TAPDevice: netAlloc.TAPDevice,
-			IP:        netAlloc.IP,
-			MAC:       netAlloc.MAC,
-			Netmask:   "255.255.255.0", // Default netmask
+			log.ErrorContext(ctx, "failed to allocate network", "instance_id", id, "error", err)
+			return nil, fmt.Errorf("allocate network: %w", err)
 		}
+		// Update stored metadata with new IP/MAC
+		stored.IP = netConfig.IP
+		stored.MAC = netConfig.MAC
+		// Add network cleanup to stack
+		cu.Add(func() {
+			m.networkManager.ReleaseAllocation(ctx, &network.Allocation{
+				InstanceID: id,
+				TAPDevice:  netConfig.TAPDevice,
+			})
+		})
 	}
 
-	// 5. Start VMM and boot VM (reuses logic from create)
+	// 5. Regenerate config disk with new network configuration
+	instForConfig := &Instance{StoredMetadata: *stored}
+	log.DebugContext(ctx, "regenerating config disk", "instance_id", id)
+	if err := m.createConfigDisk(instForConfig, imageInfo, netConfig); err != nil {
+		log.ErrorContext(ctx, "failed to create config disk", "instance_id", id, "error", err)
+		return nil, fmt.Errorf("create config disk: %w", err)
+	}
+
+	// 6. Start VMM and boot VM (reuses logic from create)
 	log.InfoContext(ctx, "starting VMM and booting VM", "instance_id", id)
 	if err := m.startAndBootVM(ctx, stored, imageInfo, netConfig); err != nil {
 		log.ErrorContext(ctx, "failed to start and boot VM", "instance_id", id, "error", err)
-		// Cleanup network on failure
-		if stored.NetworkEnabled {
-			if netAlloc, err := m.networkManager.GetAllocation(ctx, id); err == nil {
-				m.networkManager.ReleaseAllocation(ctx, netAlloc)
-			}
-		}
 		return nil, err
 	}
 
-	// 6. Update metadata (set PID, StartedAt)
+	// Success - release cleanup stack (prevent cleanup)
+	cu.Release()
+
+	// 7. Update metadata (set PID, StartedAt)
 	now := time.Now()
 	stored.StartedAt = &now