fix: libvirt backend failures (#425, #427, #430, #431, #432)

internal/api/setup/dependencies.go

@@ -247,7 +247,7 @@ func InitServices(c ServiceConfig) (*Services, *Workers, error) {
 	eventSvc := services.NewEventService(services.EventServiceParams{Repo: c.Repos.Event, RBACSvc: rbacSvc, Publisher: wsHub, Logger: c.Logger})
 
 	// 3. Cloud Infrastructure Services (VPC, Subnet, Instance, Volume, SG, LB)
-	vpcSvc := services.NewVpcService(services.VpcServiceParams{Repo: c.Repos.Vpc, LBRepo: c.Repos.LB, PeeringRepo: c.Repos.VPCPeering, AsRepo: c.Repos.AutoScaling, RBACSvc: rbacSvc, Network: c.Network, AuditSvc: auditSvc, Logger: c.Logger, DefaultCIDR: c.Config.DefaultVPCCIDR})
+	vpcSvc := services.NewVpcService(services.VpcServiceParams{Repo: c.Repos.Vpc, LBRepo: c.Repos.LB, PeeringRepo: c.Repos.VPCPeering, AsRepo: c.Repos.AutoScaling, RBACSvc: rbacSvc, Network: c.Network, AuditSvc: auditSvc, Logger: c.Logger, DefaultCIDR: c.Config.DefaultVPCCIDR, ComputeBackend: c.Config.ComputeBackend})
 	subnetSvc := services.NewSubnetService(services.SubnetServiceParams{Repo: c.Repos.Subnet, RBACSvc: rbacSvc, VpcRepo: c.Repos.Vpc, AuditSvc: auditSvc, Logger: c.Logger})
 	volumeSvc := services.NewVolumeService(services.VolumeServiceParams{Repo: c.Repos.Volume, RBACSvc: rbacSvc, Storage: c.Storage, EventSvc: eventSvc, AuditSvc: auditSvc, Logger: c.Logger})
 

internal/core/services/cache.go

@@ -174,10 +174,6 @@ func (s *CacheService) resolveNetworkID(ctx context.Context, vpcID *uuid.UUID) (
 		s.logger.Error("failed to get VPC", "vpc_id", vpcID, "error", err)
 		return "", err
 	}
-	// For Docker backend with OVS networks, don't pass br-vpc-* networks
-	if s.compute != nil && s.compute.Type() == "docker" && strings.HasPrefix(vpc.NetworkID, "br-vpc-") {
-		return "", nil
-	}
 	return vpc.NetworkID, nil
 }
 

internal/core/services/database.go

@@ -1195,9 +1195,6 @@ func (s *DatabaseService) resolveVpcNetwork(ctx context.Context, vpcID *uuid.UUI
 	if err != nil {
 		return "", err
 	}
-	if s.compute != nil && s.compute.Type() == "docker" && strings.HasPrefix(vpc.NetworkID, "br-vpc-") {
-		return "", nil
-	}
 	return vpc.NetworkID, nil
 }
 

internal/core/services/instance.go

@@ -606,8 +606,8 @@ func (s *InstanceService) StopInstance(ctx context.Context, idOrName string) err
 
 	if err := s.compute.StopInstance(ctx, target); err != nil {
 		platform.InstanceOperationsTotal.WithLabelValues("stop", "failure").Inc()
-		s.logger.Error("failed to stop docker container", "container_id", target, "error", err)
-		return errors.Wrap(errors.Internal, "failed to stop container", err)
+		s.logger.Error("failed to stop instance", "backend", s.compute.Type(), "id", target, "error", err)
+		return errors.Wrap(errors.Internal, fmt.Sprintf("failed to stop %s instance", s.compute.Type()), err)
 	}
 
 	platform.InstancesTotal.WithLabelValues("running", s.compute.Type()).Dec()

internal/core/services/vpc.go

@@ -29,6 +29,9 @@ type VpcServiceParams struct {
 	AuditSvc       ports.AuditService
 	Logger         *slog.Logger
 	DefaultCIDR    string
+	// ComputeBackend is the compute backend type ("docker", "libvirt", "firecracker").
+	// When "libvirt", VPC creation skips OVS bridge creation since libvirt manages its own networking.
+	ComputeBackend string
 }
 
 // VpcService handles the lifecycle of Virtual Private Clouds (VPCs),
@@ -44,6 +47,7 @@ type VpcService struct {
 	auditSvc       ports.AuditService
 	logger         *slog.Logger
 	defaultCIDR    string
+	computeBackend string
 }
 
 // NewVpcService creates a new instance of VpcService.
@@ -68,6 +72,7 @@ func NewVpcService(params VpcServiceParams) *VpcService {
 		auditSvc:       params.AuditSvc,
 		logger:         logger,
 		defaultCIDR:    defaultCIDR,
+		computeBackend: params.ComputeBackend,
 	}
 }
 
@@ -102,10 +107,19 @@ func (s *VpcService) CreateVPC(ctx context.Context, name, cidrBlock string) (*do
 	// 1. Generate unique VNI (for demo purposes we use a hash based int)
 	vxlanID := int(vpcID[0]) + 100
 
-	// 2. Create OVS bridge
-	bridgeName := fmt.Sprintf("br-vpc-%s", vpcID.String()[:8])
-	if err := s.network.CreateBridge(ctx, bridgeName, vxlanID); err != nil {
-		return nil, errors.Wrap(errors.Internal, "failed to create OVS bridge", err)
+	var bridgeName string
+	var bridgeCreated bool
+
+	// 2. Create OVS bridge (skipped for libvirt compute backend)
+	if s.computeBackend != "libvirt" {
+		if s.network == nil {
+			return nil, errors.New(errors.Internal, "network backend is required for VPC bridge creation")
+		}
+		bridgeName = fmt.Sprintf("br-vpc-%s", vpcID.String()[:8])
+		if err := s.network.CreateBridge(ctx, bridgeName, vxlanID); err != nil {
+			return nil, errors.Wrap(errors.Internal, "failed to create OVS bridge", err)
+		}
+		bridgeCreated = true
 	}
 
 	// 3. Construct ARN
@@ -118,7 +132,7 @@ func (s *VpcService) CreateVPC(ctx context.Context, name, cidrBlock string) (*do
 		TenantID:  tenantID,
 		Name:      name,
 		CIDRBlock: cidrBlock,
-		NetworkID: bridgeName,
+		NetworkID: bridgeName, // empty string for libvirt
 		VXLANID:   vxlanID,
 		Status:    "active",
 		ARN:       arn,
@@ -127,9 +141,11 @@ func (s *VpcService) CreateVPC(ctx context.Context, name, cidrBlock string) (*do
 
 	if err := s.repo.Create(ctx, vpc); err != nil {
 		// Cleanup OVS bridge if DB fails
-		s.logger.Error("failed to create VPC in DB, rolling back bridge", "name", name, "error", err)
-		if rbErr := s.network.DeleteBridge(ctx, bridgeName); rbErr != nil {
-			s.logger.Error("failed to rollback bridge", "bridge", bridgeName, "error", rbErr)
+		if bridgeCreated {
+			s.logger.Error("failed to create VPC in DB, rolling back bridge", "name", name, "error", err)
+			if rbErr := s.network.DeleteBridge(ctx, bridgeName); rbErr != nil {
+				s.logger.Error("failed to rollback bridge", "bridge", bridgeName, "error", rbErr)
+			}
 		}
 		return nil, errors.Wrap(errors.Internal, "failed to create VPC in database", err)
 	}
@@ -153,7 +169,9 @@ func (s *VpcService) CreateVPC(ctx context.Context, name, cidrBlock string) (*do
 			// Rollback: delete VPC
 			s.logger.Error("failed to create main route table, rolling back VPC", "error", err)
 			_ = s.repo.Delete(ctx, vpc.ID)
-			_ = s.network.DeleteBridge(ctx, bridgeName)
+			if bridgeCreated {
+				_ = s.network.DeleteBridge(ctx, bridgeName)
+			}
 			return nil, errors.Wrap(errors.Internal, "failed to create main route table", err)
 		}
 	}
@@ -224,12 +242,14 @@ func (s *VpcService) DeleteVPC(ctx context.Context, idOrName string, force bool)
 		}
 	}
 
-	// 2. Remove OVS bridge
-	if err := s.network.DeleteBridge(ctx, vpc.NetworkID); err != nil {
-		s.logger.Error("failed to remove OVS bridge", "bridge", vpc.NetworkID, "error", err)
-		return errors.Wrap(errors.Internal, "failed to remove OVS bridge", err)
+	// 2. Remove OVS bridge (skip for libvirt where NetworkID is empty)
+	if vpc.NetworkID != "" {
+		if err := s.network.DeleteBridge(ctx, vpc.NetworkID); err != nil {
+			s.logger.Error("failed to remove OVS bridge", "bridge", vpc.NetworkID, "error", err)
+			return errors.Wrap(errors.Internal, "failed to remove OVS bridge", err)
+		}
+		s.logger.Info("vpc bridge removed", "bridge", vpc.NetworkID)
 	}
-	s.logger.Info("vpc bridge removed", "bridge", vpc.NetworkID)
 
 	// 3. Delete from DB
 	if err := s.repo.Delete(ctx, vpc.ID); err != nil {

internal/platform/resilient_compute.go

@@ -54,6 +54,12 @@ func (o ResilientComputeOpts) withDefaults() ResilientComputeOpts {
 
 // ResilientCompute wraps a ComputeBackend with circuit breaker, bulkhead,
 // and per-call timeouts. It implements the ports.ComputeBackend interface.
+//
+// Note: the circuit breaker is per-backend-instance (one breaker guards all
+// operations on a single compute backend). If one operation fails repeatedly,
+// the breaker trips and blocks all subsequent calls to that backend until the
+// reset timeout expires. This is a deliberate design trade-off for simplicity.
+// For finer-grained isolation, each operation type could have its own breaker.
 type ResilientCompute struct {
 	inner    ports.ComputeBackend
 	cb       *CircuitBreaker