chore/inspektor_gadget: Add upgrade action

docs/inspektor-gadget-usage.md

@@ -67,7 +67,7 @@ Can you observe system calls for the pod my-pod in the default namespace for few
 
 ## Prerequisites
 
-- A kubeconfig file that has access to the AKS cluster. You will need to restart the MCP server if you change the kubeconfig file.
+- A kubeconfig file that has access to the AKS cluster.
 - The tool requires Inspektor Gadget to be installed in the cluster. If you are running with `--access-level=readwrite` or more, the MCP server will automatically
   install Inspektor Gadget (action `deploy` ) in the cluster otherwise you can follow the steps to install it manually: [Inspektor Gadget Installation](https://learn.microsoft.com/en-us/troubleshoot/azure/azure-kubernetes/logs/capture-system-insights-from-aks#how-to-install-inspektor-gadget-in-an-aks-cluster) or
   use the official Helm chart: [Inspektor Gadget Helm Chart](https://inspektor-gadget.io/docs/latest/reference/install-kubernetes#installation-with-the-helm-chart):

internal/components/inspektorgadget/const.go

@@ -6,6 +6,8 @@ const (
 	deployAction = "deploy"
 	// undeployAction is the action to remove Inspektor Gadget from the cluster
 	undeployAction = "undeploy"
+	// upgradeAction is the action to upgrade Inspektor Gadget in the cluster
+	upgradeAction = "upgrade"
 	// isDeployedAction is the action to check if Inspektor Gadget is deployed
 	isDeployedAction = "is_deployed"
 )

internal/components/inspektorgadget/gadgetmanager.go

@@ -7,6 +7,7 @@ import (
 	"fmt"
 	"slices"
 	"strings"
+	"sync"
 	"time"
 
 	"github.com/inspektor-gadget/inspektor-gadget/pkg/datasource"
@@ -15,7 +16,6 @@ import (
 	gadgetcontext "github.com/inspektor-gadget/inspektor-gadget/pkg/gadget-context"
 	"github.com/inspektor-gadget/inspektor-gadget/pkg/operators"
 	"github.com/inspektor-gadget/inspektor-gadget/pkg/operators/simple"
-	"github.com/inspektor-gadget/inspektor-gadget/pkg/runtime"
 	grpcruntime "github.com/inspektor-gadget/inspektor-gadget/pkg/runtime/grpc"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
 	"k8s.io/cli-runtime/pkg/genericclioptions"
@@ -40,8 +40,8 @@ type GadgetManager interface {
 	ListGadgets(ctx context.Context) ([]*GadgetInstance, error)
 	// IsDeployed checks if the Inspektor Gadget is deployed in the environment
 	IsDeployed(ctx context.Context) (bool, string, error)
-	// Close closes the gadget manager and releases any resources
-	Close() error
+	// GetVersion retrieves the version of Inspektor Gadget installed in the cluster
+	GetVersion() (string, error)
 }
 
 // GadgetInstance represents a running gadget instance
@@ -60,25 +60,12 @@ func init() {
 }
 
 // NewGadgetManager creates a new instance of GadgetManager
-func NewGadgetManager() (GadgetManager, error) {
-	rt := grpcruntime.New(grpcruntime.WithConnectUsingK8SProxy)
-	if err := rt.Init(nil); err != nil {
-		return nil, fmt.Errorf("initializing gadget runtime: %w", err)
-	}
-
-	restConfig, err := KubernetesFlags.ToRESTConfig()
-	if err != nil {
-		return nil, fmt.Errorf("creating REST config: %w", err)
-	}
-	rt.SetRestConfig(restConfig)
-
-	return &manager{
-		runtime: rt,
-	}, nil
+func NewGadgetManager() GadgetManager {
+	return &manager{}
 }
 
 type manager struct {
-	runtime runtime.Runtime
+	formatterMu sync.Mutex
 }
 
 // RunGadget runs a gadget with the specified image and parameters for a given duration
@@ -88,15 +75,20 @@ func (g *manager) RunGadget(ctx context.Context, image string, params map[string
 		ctx,
 		image,
 		gadgetcontext.WithDataOperators(
-			outputDataOperator(func(data []byte) {
+			g.outputDataOperator(func(data []byte) {
 				results.Write(data)
 				results.WriteByte('\n')
 			}),
 		),
 		gadgetcontext.WithTimeout(duration),
 	)
 
-	if err := g.runtime.RunGadget(gadgetCtx, g.runtime.ParamDescs().ToParams(), params); err != nil {
+	rt, err := getRuntime()
+	if err != nil {
+		return "", fmt.Errorf("getting runtime: %w", err)
+	}
+
+	if err := rt.RunGadget(gadgetCtx, rt.ParamDescs().ToParams(), params); err != nil {
 		return "", fmt.Errorf("running gadget: %w", err)
 	}
 
@@ -118,7 +110,7 @@ func truncateResults(results string, latest bool) string {
 	return fmt.Sprintf("\n<isTruncated>true</isTruncated>\n<results>%s</results>\n", truncated)
 }
 
-func outputDataOperator(cb func(data []byte)) operators.DataOperator {
+func (g *manager) outputDataOperator(cb func(data []byte)) operators.DataOperator {
 	const opPriority = 50000
 	return simple.New("outputDataOperator",
 		simple.OnInit(func(gadgetCtx operators.GadgetContext) error {
@@ -135,6 +127,8 @@ func outputDataOperator(cb func(data []byte)) operators.DataOperator {
 				}
 
 				err := d.Subscribe(func(source datasource.DataSource, data datasource.Data) error {
+					g.formatterMu.Lock()
+					defer g.formatterMu.Unlock()
 					jsonData := jsonFormatter.Marshal(data)
 					cb(jsonData)
 					return nil
@@ -155,10 +149,14 @@ func (g *manager) StartGadget(ctx context.Context, image string, params map[stri
 		image,
 	)
 
-	p := g.runtime.ParamDescs().ToParams()
+	rt, err := getRuntime()
+	if err != nil {
+		return "", fmt.Errorf("getting runtime: %w", err)
+	}
+	p := rt.ParamDescs().ToParams()
 
 	newID := make([]byte, 16)
-	_, err := rand.Read(newID)
+	_, err = rand.Read(newID)
 	if err != nil {
 		return "", fmt.Errorf("generating new gadget ID: %w", err)
 	}
@@ -175,7 +173,7 @@ func (g *manager) StartGadget(ctx context.Context, image string, params map[stri
 	if err = p.Set(grpcruntime.ParamTags, strings.Join(append(tags, "createdBy=aks-mcp"), ",")); err != nil {
 		return "", fmt.Errorf("setting gadget tags: %w", err)
 	}
-	if err := g.runtime.RunGadget(gadgetCtx, p, params); err != nil {
+	if err = rt.RunGadget(gadgetCtx, p, params); err != nil {
 		return "", fmt.Errorf("running gadget: %w", err)
 	}
 
@@ -184,8 +182,13 @@ func (g *manager) StartGadget(ctx context.Context, image string, params map[stri
 
 // StopGadget stops a running gadget by its ID
 func (g *manager) StopGadget(ctx context.Context, id string) error {
-	if err := g.runtime.(*grpcruntime.Runtime).RemoveGadgetInstance(ctx, g.runtime.ParamDescs().ToParams(), id); err != nil {
-		return fmt.Errorf("stopping to gadget: %w", err)
+	rt, err := getRuntime()
+	if err != nil {
+		return fmt.Errorf("getting runtime: %w", err)
+	}
+
+	if err = rt.RemoveGadgetInstance(ctx, rt.ParamDescs().ToParams(), id); err != nil {
+		return fmt.Errorf("stopping gadget: %w", err)
 	}
 	return nil
 }
@@ -200,7 +203,7 @@ func (g *manager) GetResults(ctx context.Context, id string) (string, error) {
 		to,
 		id,
 		gadgetcontext.WithDataOperators(
-			outputDataOperator(func(data []byte) {
+			g.outputDataOperator(func(data []byte) {
 				results.Write(data)
 				results.WriteByte('\n')
 			}),
@@ -210,7 +213,12 @@ func (g *manager) GetResults(ctx context.Context, id string) (string, error) {
 		gadgetcontext.WithTimeout(time.Second),
 	)
 
-	if err := g.runtime.RunGadget(gadgetCtx, g.runtime.ParamDescs().ToParams(), map[string]string{}); err != nil {
+	rt, err := getRuntime()
+	if err != nil {
+		return "", fmt.Errorf("getting runtime: %w", err)
+	}
+
+	if err = rt.RunGadget(gadgetCtx, rt.ParamDescs().ToParams(), map[string]string{}); err != nil {
 		return "", fmt.Errorf("attaching to gadget: %w", err)
 	}
 
@@ -219,7 +227,12 @@ func (g *manager) GetResults(ctx context.Context, id string) (string, error) {
 
 // ListGadgets lists all running gadgets and returns their instances
 func (g *manager) ListGadgets(ctx context.Context) ([]*GadgetInstance, error) {
-	instances, err := g.runtime.(*grpcruntime.Runtime).GetGadgetInstances(ctx, g.runtime.ParamDescs().ToParams())
+	rt, err := getRuntime()
+	if err != nil {
+		return nil, fmt.Errorf("getting runtime: %w", err)
+	}
+
+	instances, err := rt.GetGadgetInstances(ctx, rt.ParamDescs().ToParams())
 	if err != nil {
 		return nil, fmt.Errorf("listing gadgets: %w", err)
 	}
@@ -267,10 +280,31 @@ func (g *manager) IsDeployed(ctx context.Context) (bool, string, error) {
 	return true, namespaces[0], nil
 }
 
-// Close closes the gadget manager and releases any resources
-func (g *manager) Close() error {
-	if g.runtime != nil {
-		return g.runtime.Close()
+func (g *manager) GetVersion() (string, error) {
+	rt, err := getRuntime()
+	if err != nil {
+		return "", fmt.Errorf("getting runtime: %w", err)
 	}
-	return nil
+
+	info, err := rt.GetInfo()
+	if err != nil {
+		return "", fmt.Errorf("getting info: %w", err)
+	}
+	return info.ServerVersion, nil
+}
+
+// getRuntime sets up a runtime, ensuring we always use the latest kubeconfig
+func getRuntime() (*grpcruntime.Runtime, error) {
+	rt := grpcruntime.New(grpcruntime.WithConnectUsingK8SProxy)
+	if err := rt.Init(nil); err != nil {
+		return nil, fmt.Errorf("initializing gadget runtime: %w", err)
+	}
+
+	restConfig, err := KubernetesFlags.ToRESTConfig()
+	if err != nil {
+		return nil, fmt.Errorf("creating REST config: %w", err)
+	}
+	rt.SetRestConfig(restConfig)
+
+	return rt, nil
 }

internal/components/inspektorgadget/gadgets.go

@@ -19,10 +19,14 @@ type Gadget struct {
 	ParamsFunc func(filterParams map[string]interface{}, gadgetParams map[string]string)
 }
 
+func (g *Gadget) getImage(version string) string {
+	return fmt.Sprintf("%s:%s", g.Image, gadgetVersionFor(version))
+}
+
 var gadgets = []Gadget{
 	{
 		Name:        observeDNS,
-		Image:       "ghcr.io/inspektor-gadget/gadget/trace_dns:latest",
+		Image:       "ghcr.io/inspektor-gadget/gadget/trace_dns",
 		Description: "Observes DNS queries in the cluster",
 		Params: map[string]interface{}{
 			"name": map[string]interface{}{
@@ -75,7 +79,7 @@ var gadgets = []Gadget{
 	},
 	{
 		Name:        observeTCP,
-		Image:       "ghcr.io/inspektor-gadget/gadget/trace_tcp:latest",
+		Image:       "ghcr.io/inspektor-gadget/gadget/trace_tcp",
 		Description: "Observes TCP traffic in the cluster",
 		Params: map[string]interface{}{
 			"source_port": map[string]interface{}{
@@ -121,7 +125,7 @@ var gadgets = []Gadget{
 	},
 	{
 		Name:        observeFileOpen,
-		Image:       "ghcr.io/inspektor-gadget/gadget/trace_open:latest",
+		Image:       "ghcr.io/inspektor-gadget/gadget/trace_open",
 		Description: "Observes file open operations in the cluster",
 		Params: map[string]interface{}{
 			"path": map[string]interface{}{
@@ -152,7 +156,7 @@ var gadgets = []Gadget{
 	},
 	{
 		Name:        observeProcessExecution,
-		Image:       "ghcr.io/inspektor-gadget/gadget/trace_exec:latest",
+		Image:       "ghcr.io/inspektor-gadget/gadget/trace_exec",
 		Description: "Observes process execution in the cluster",
 		Params: map[string]interface{}{
 			"command": map[string]interface{}{
@@ -172,7 +176,7 @@ var gadgets = []Gadget{
 	},
 	{
 		Name:        observeSignal,
-		Image:       "ghcr.io/inspektor-gadget/gadget/trace_signal:latest",
+		Image:       "ghcr.io/inspektor-gadget/gadget/trace_signal",
 		Description: "Traces signals sent to containers in the cluster",
 		Params: map[string]interface{}{
 			"signal": map[string]interface{}{
@@ -193,7 +197,7 @@ var gadgets = []Gadget{
 	},
 	{
 		Name:        observeSystemCalls,
-		Image:       "ghcr.io/inspektor-gadget/gadget/traceloop:latest",
+		Image:       "ghcr.io/inspektor-gadget/gadget/traceloop",
 		Description: "Observes system calls in the cluster",
 		Params: map[string]interface{}{
 			"syscall": map[string]interface{}{
@@ -215,7 +219,7 @@ var gadgets = []Gadget{
 	},
 	{
 		Name:        topFile,
-		Image:       "ghcr.io/inspektor-gadget/gadget/top_file:latest",
+		Image:       "ghcr.io/inspektor-gadget/gadget/top_file",
 		Description: "Shows top files by read/write operations",
 		Params: map[string]interface{}{
 			"max_entries": map[string]interface{}{
@@ -240,7 +244,7 @@ var gadgets = []Gadget{
 	},
 	{
 		Name:        topTCP,
-		Image:       "ghcr.io/inspektor-gadget/gadget/top_tcp:latest",
+		Image:       "ghcr.io/inspektor-gadget/gadget/top_tcp",
 		Description: "Shows top TCP connections by traffic volume",
 		Params: map[string]interface{}{
 			"max_entries": map[string]interface{}{

internal/components/inspektorgadget/gadgets_test.go

@@ -1,6 +1,9 @@
 package inspektorgadget
 
-import "testing"
+import (
+	"strings"
+	"testing"
+)
 
 func TestGadgets(t *testing.T) {
 	for _, gadget := range gadgets {
@@ -10,6 +13,9 @@ func TestGadgets(t *testing.T) {
 		if gadget.Image == "" {
 			t.Errorf("Gadget image is empty for %s", gadget.Name)
 		}
+		if gadget.Image != "" && strings.Contains(gadget.Image, ":") {
+			t.Errorf("Gadget image %s should not contain a version tag", gadget.Image)
+		}
 		if gadget.Description == "" {
 			t.Errorf("Gadget description is empty for %s", gadget.Name)
 		}