Pass rollback window duration to upgrade watcher command

internal/pkg/agent/application/application.go

@@ -118,7 +118,7 @@ func New(
 
 	// monitoring is not supported in bootstrap mode https://github.com/elastic/elastic-agent/issues/1761
 	isMonitoringSupported := !disableMonitoring && cfg.Settings.V1MonitoringEnabled
-	upgrader, err := upgrade.NewUpgrader(log, cfg.Settings.DownloadConfig, agentInfo)
+	upgrader, err := upgrade.NewUpgrader(log, cfg.Settings.DownloadConfig, cfg.Settings.Upgrade, agentInfo)
 	if err != nil {
 		return nil, nil, nil, fmt.Errorf("failed to create upgrader: %w", err)
 	}

internal/pkg/agent/application/coordinator/coordinator_unit_test.go

@@ -35,6 +35,7 @@ import (
 	"github.com/elastic/elastic-agent/internal/pkg/agent/application/upgrade"
 	"github.com/elastic/elastic-agent/internal/pkg/agent/application/upgrade/artifact"
 	"github.com/elastic/elastic-agent/internal/pkg/agent/application/upgrade/details"
+	"github.com/elastic/elastic-agent/internal/pkg/agent/configuration"
 	"github.com/elastic/elastic-agent/internal/pkg/agent/transpiler"
 	"github.com/elastic/elastic-agent/internal/pkg/config"
 	monitoringCfg "github.com/elastic/elastic-agent/internal/pkg/core/monitoring/config"
@@ -446,6 +447,7 @@ func TestCoordinatorReportsInvalidPolicy(t *testing.T) {
 	upgradeMgr, err := upgrade.NewUpgrader(
 		log,
 		&artifact.Config{},
+		&configuration.UpgradeConfig{},
 		&info.AgentInfo{},
 	)
 	require.NoError(t, err, "errored when creating a new upgrader")

internal/pkg/agent/application/upgrade/rollback.go

@@ -24,6 +24,7 @@
 	"github.com/elastic/elastic-agent/pkg/control/v2/client"
 	"github.com/elastic/elastic-agent/pkg/core/logger"
 	"github.com/elastic/elastic-agent/pkg/utils"
+	"github.com/elastic/elastic-agent/pkg/version"
 )
 
 const (
@@ -33,6 +34,9 @@
 	restartBackoffMax  = 90 * time.Second
 )
 
+// Rollback window feature is only available starting with version >= 9.1.0-SNAPSHOT.
+var rollbackWindowMinVersion = version.NewParsedSemVer(9, 1, 0, "SNAPSHOT", "")
+
 // Rollback rollbacks to previous version which was functioning before upgrade.
 func Rollback(ctx context.Context, log *logger.Logger, c client.Client, topDirPath, prevVersionedHome, prevHash string) error {
 	symlinkPath := filepath.Join(topDirPath, agentName)
@@ -144,13 +148,13 @@
 
 // InvokeWatcher invokes an agent instance using watcher argument for watching behavior of
 // agent during upgrade period.
-func InvokeWatcher(log *logger.Logger, agentExecutable string) (*exec.Cmd, error) {
+func InvokeWatcher(log *logger.Logger, agentExecutable string, rollbackWindow time.Duration) (*exec.Cmd, error) {
 	if !IsUpgradeable() {
 		log.Info("agent is not upgradable, not starting watcher")
 		return nil, nil
 	}
 
-	cmd := invokeCmd(agentExecutable)
+	cmd := makeOSWatchCmd(makeBaseWatchCmd(agentExecutable, rollbackWindow))
 	log.Infow("Starting upgrade watcher", "path", cmd.Path, "args", cmd.Args, "env", cmd.Env, "dir", cmd.Dir)
 	if err := cmd.Start(); err != nil {
 		return nil, fmt.Errorf("failed to start Upgrade Watcher: %w", err)
@@ -238,3 +242,24 @@
 	close(signal)
 	return nil
 }
+
+func makeBaseWatchCmd(agentExecutable string, rollbackWindow time.Duration) *exec.Cmd {
+	cmdArgs := []string{
+		watcherSubcommand,
+		"--path.config", paths.Config(),
+		"--path.home", paths.Top(),
+	}
+
+	if rollbackWindow > 0 {
+		cmdArgs = append(cmdArgs, "--rollback-window", fmt.Sprintf("%s", rollbackWindow.String()))
+	}
+
+	// #nosec G204 -- user cannot inject any parameters to this command
+	return exec.Command(agentExecutable, cmdArgs...)
+}
+
+// isRollbackWindowSupported checks if the rollback window feature is supported by the version of
+// the target Agent, i.e the version being upgraded *to*.
+func isRollbackWindowSupported(targetAgentVersion *version.ParsedSemVer) bool {
+	return !targetAgentVersion.Less(*rollbackWindowMinVersion)
+}

internal/pkg/agent/application/upgrade/rollback_darwin.go

@@ -11,8 +11,6 @@ import (
 	"os/exec"
 	"syscall"
 	"time"
-
-	"github.com/elastic/elastic-agent/internal/pkg/agent/application/paths"
 )
 
 const (
@@ -21,13 +19,7 @@ const (
 	afterRestartDelay = 2 * time.Second
 )
 
-func invokeCmd(agentExecutable string) *exec.Cmd {
-	// #nosec G204 -- user cannot inject any parameters to this command
-	cmd := exec.Command(agentExecutable, watcherSubcommand,
-		"--path.config", paths.Config(),
-		"--path.home", paths.Top(),
-	)
-
+func makeOSWatchCmd(baseWatchCmd *exec.Cmd) *exec.Cmd {
 	var cred = &syscall.Credential{
 		Uid:         uint32(os.Getuid()),
 		Gid:         uint32(os.Getgid()),
@@ -38,6 +30,6 @@ func invokeCmd(agentExecutable string) *exec.Cmd {
 		Credential: cred,
 		Setsid:     true,
 	}
-	cmd.SysProcAttr = sysproc
-	return cmd
+	baseWatchCmd.SysProcAttr = sysproc
+	return baseWatchCmd
 }

internal/pkg/agent/application/upgrade/rollback_linux.go

@@ -11,8 +11,6 @@
 	"os/exec"
 	"syscall"
 	"time"
-
-	"github.com/elastic/elastic-agent/internal/pkg/agent/application/paths"
 )
 
 const (
@@ -21,16 +19,10 @@
 	afterRestartDelay = 2 * time.Second
 )
 
-func invokeCmd(agentExecutable string) *exec.Cmd {
-	// #nosec G204 -- user cannot inject any parameters to this command
-	cmd := exec.Command(agentExecutable, watcherSubcommand,
-		"--path.config", paths.Config(),
-		"--path.home", paths.Top(),
-	)
-
+func makeOSWatchCmd(baseWatchCmd *exec.Cmd) *exec.Cmd {
 	var cred = &syscall.Credential{
 		Uid:         uint32(os.Getuid()),
 		Gid:         uint32(os.Getgid()),
 		Groups:      nil,
 		NoSetGroups: true,
 	}
@@ -40,6 +32,6 @@
 		// propagate sigint instead of sigkill so we can ignore it
 		Pdeathsig: syscall.SIGINT,
 	}
-	cmd.SysProcAttr = sysproc
-	return cmd
+	baseWatchCmd.SysProcAttr = sysproc
+	return baseWatchCmd
 }

internal/pkg/agent/application/upgrade/rollback_test.go

@@ -11,6 +11,7 @@ import (
 	"path/filepath"
 	"runtime"
 	"testing"
+	"time"
 
 	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/mock"
@@ -19,6 +20,7 @@ import (
 	"github.com/elastic/elastic-agent/internal/pkg/agent/application/paths"
 	"github.com/elastic/elastic-agent/pkg/core/logger"
 	"github.com/elastic/elastic-agent/pkg/core/logger/loggertest"
+	"github.com/elastic/elastic-agent/pkg/version"
 	mocks "github.com/elastic/elastic-agent/testing/mocks/pkg/control/v2/client"
 )
 
@@ -328,6 +330,78 @@ func TestRollback(t *testing.T) {
 	}
 }
 
+func TestIsRollbackWindowSupported(t *testing.T) {
+	tests := map[string]struct {
+		version string
+		want    bool
+	}{
+		"supported_version": {
+			version: "9.2.0-SNAPSHOT",
+			want:    true,
+		},
+		"older_version": {
+			version: "9.0.0-SNAPSHOT",
+			want:    false,
+		},
+		"exactly_minimum_version": {
+			version: "9.1.0",
+			want:    true,
+		},
+	}
+
+	for name, tt := range tests {
+		t.Run(name, func(t *testing.T) {
+			v, err := version.ParseVersion(tt.version)
+			require.NoError(t, err)
+
+			got := isRollbackWindowSupported(v)
+			assert.Equal(t, tt.want, got, "isRollbackWindowSupported(%q)", tt.version)
+		})
+	}
+}
+
+func TestMakeBaseWatchCmd(t *testing.T) {
+	exec, err := os.Executable()
+	require.NoError(t, err)
+	exec, err = filepath.EvalSymlinks(exec)
+	require.NoError(t, err)
+	execDir := filepath.Dir(exec)
+
+	agentExecutable := "elastic-agent"
+	t.Run("no_rollback_window", func(t *testing.T) {
+		cmd := makeBaseWatchCmd(agentExecutable, 0)
+
+		// Expected command:
+		// elastic-agent watch --path.config /some/path --path.home /some/path
+		require.Equal(t, cmd.Args, []string{
+			agentExecutable,
+			"watch",
+			"--path.config",
+			execDir,
+			"--path.home",
+			execDir,
+		})
+	})
+
+	t.Run("with_rollback_window", func(t *testing.T) {
+		rollbackWindow := 2*time.Hour + 15*time.Minute
+		cmd := makeBaseWatchCmd(agentExecutable, rollbackWindow)
+
+		// Expected command:
+		// elastic-agent watch --path.config /some/path --path.home /some/path --rollback-window 8100s
+		require.Equal(t, cmd.Args, []string{
+			agentExecutable,
+			"watch",
+			"--path.config",
+			execDir,
+			"--path.home",
+			execDir,
+			"--rollback-window",
+			rollbackWindow.String(),
+		})
+	})
+}
+
 // checkFilesAfterCleanup is a convenience function to check the file structure within topDir.
 // *AgentHome paths must be the expected old and new agent paths relative to topDir (typically in the form of "data/elastic-agent-*")
 func checkFilesAfterCleanup(t *testing.T, topDir, newAgentHome string, oldAgentHomes ...string) {

internal/pkg/agent/application/upgrade/rollback_windows.go

@@ -9,8 +9,6 @@ package upgrade
 import (
 	"os/exec"
 	"time"
-
-	"github.com/elastic/elastic-agent/internal/pkg/agent/application/paths"
 )
 
 const (
@@ -19,11 +17,6 @@ const (
 	afterRestartDelay = 20 * time.Second
 )
 
-func invokeCmd(agentExecutable string) *exec.Cmd {
-	// #nosec G204 -- user cannot inject any parameters to this command
-	cmd := exec.Command(agentExecutable, watcherSubcommand,
-		"--path.config", paths.Config(),
-		"--path.home", paths.Top(),
-	)
-	return cmd
+func makeOSWatchCmd(baseWatchCmd *exec.Cmd) *exec.Cmd {
+	return baseWatchCmd
 }

internal/pkg/agent/application/upgrade/step_download.go

@@ -50,7 +50,7 @@ func (u *Upgrader) downloadArtifact(ctx context.Context, parsedVersion *agtversi
 	pgpBytes = u.appendFallbackPGP(parsedVersion, pgpBytes)
 
 	// do not update source config
-	settings := *u.settings
+	settings := *u.downloadSettings
 	var downloaderFunc downloader
 	var factory downloaderFactory
 	var verifier download.Verifier

internal/pkg/agent/application/upgrade/step_download_test.go

@@ -19,6 +19,7 @@ import (
 	"github.com/elastic/elastic-agent/internal/pkg/agent/application/upgrade/artifact"
 	"github.com/elastic/elastic-agent/internal/pkg/agent/application/upgrade/artifact/download"
 	"github.com/elastic/elastic-agent/internal/pkg/agent/application/upgrade/details"
+	"github.com/elastic/elastic-agent/internal/pkg/agent/configuration"
 	"github.com/elastic/elastic-agent/internal/pkg/agent/errors"
 	"github.com/elastic/elastic-agent/pkg/core/logger"
 	"github.com/elastic/elastic-agent/pkg/core/logger/loggertest"
@@ -78,29 +79,30 @@ func TestDownloadWithRetries(t *testing.T) {
 	expectedDownloadPath := "https://artifacts.elastic.co/downloads/beats/elastic-agent"
 	testLogger, obs := loggertest.New("TestDownloadWithRetries")
 
-	settings := artifact.Config{
+	downloadSettings := artifact.Config{
 		RetrySleepInitDuration: 20 * time.Millisecond,
 		HTTPTransportSettings: httpcommon.HTTPTransportSettings{
 			Timeout: 2 * time.Second,
 		},
 	}
+	upgradeSettings := configuration.UpgradeConfig{}
 
 	// Successful immediately (no retries)
 	t.Run("successful_immediately", func(t *testing.T) {
 		mockDownloaderCtor := func(version *agtversion.ParsedSemVer, log *logger.Logger, settings *artifact.Config, upgradeDetails *details.Details) (download.Downloader, error) {
 			return &mockDownloader{expectedDownloadPath, nil}, nil
 		}
 
-		u, err := NewUpgrader(testLogger, &settings, &info.AgentInfo{})
+		u, err := NewUpgrader(testLogger, &downloadSettings, &upgradeSettings, &info.AgentInfo{})
 		require.NoError(t, err)
 
 		parsedVersion, err := agtversion.ParseVersion("8.9.0")
 		require.NoError(t, err)
 
 		upgradeDetails, upgradeDetailsRetryUntil, upgradeDetailsRetryUntilWasUnset, upgradeDetailsRetryErrorMsg := mockUpgradeDetails(parsedVersion)
-		minRetryDeadline := time.Now().Add(settings.Timeout)
+		minRetryDeadline := time.Now().Add(downloadSettings.Timeout)
 
-		path, err := u.downloadWithRetries(context.Background(), mockDownloaderCtor, parsedVersion, &settings, upgradeDetails)
+		path, err := u.downloadWithRetries(context.Background(), mockDownloaderCtor, parsedVersion, &downloadSettings, upgradeDetails)
 		require.NoError(t, err)
 		require.Equal(t, expectedDownloadPath, path)
 
@@ -141,16 +143,16 @@ func TestDownloadWithRetries(t *testing.T) {
 			return nil, nil
 		}
 
-		u, err := NewUpgrader(testLogger, &settings, &info.AgentInfo{})
+		u, err := NewUpgrader(testLogger, &downloadSettings, &upgradeSettings, &info.AgentInfo{})
 		require.NoError(t, err)
 
 		parsedVersion, err := agtversion.ParseVersion("8.9.0")
 		require.NoError(t, err)
 
 		upgradeDetails, upgradeDetailsRetryUntil, upgradeDetailsRetryUntilWasUnset, upgradeDetailsRetryErrorMsg := mockUpgradeDetails(parsedVersion)
-		minRetryDeadline := time.Now().Add(settings.Timeout)
+		minRetryDeadline := time.Now().Add(downloadSettings.Timeout)
 
-		path, err := u.downloadWithRetries(context.Background(), mockDownloaderCtor, parsedVersion, &settings, upgradeDetails)
+		path, err := u.downloadWithRetries(context.Background(), mockDownloaderCtor, parsedVersion, &downloadSettings, upgradeDetails)
 		require.NoError(t, err)
 		require.Equal(t, expectedDownloadPath, path)
 
@@ -196,16 +198,16 @@ func TestDownloadWithRetries(t *testing.T) {
 			return nil, nil
 		}
 
-		u, err := NewUpgrader(testLogger, &settings, &info.AgentInfo{})
+		u, err := NewUpgrader(testLogger, &downloadSettings, &upgradeSettings, &info.AgentInfo{})
 		require.NoError(t, err)
 
 		parsedVersion, err := agtversion.ParseVersion("8.9.0")
 		require.NoError(t, err)
 
 		upgradeDetails, upgradeDetailsRetryUntil, upgradeDetailsRetryUntilWasUnset, upgradeDetailsRetryErrorMsg := mockUpgradeDetails(parsedVersion)
-		minRetryDeadline := time.Now().Add(settings.Timeout)
+		minRetryDeadline := time.Now().Add(downloadSettings.Timeout)
 
-		path, err := u.downloadWithRetries(context.Background(), mockDownloaderCtor, parsedVersion, &settings, upgradeDetails)
+		path, err := u.downloadWithRetries(context.Background(), mockDownloaderCtor, parsedVersion, &downloadSettings, upgradeDetails)
 		require.NoError(t, err)
 		require.Equal(t, expectedDownloadPath, path)
 
@@ -231,7 +233,7 @@ func TestDownloadWithRetries(t *testing.T) {
 
 	// Download timeout expired (before all retries are exhausted)
 	t.Run("download_timeout_expired", func(t *testing.T) {
-		testCaseSettings := settings
+		testCaseSettings := downloadSettings
 		testCaseSettings.Timeout = 500 * time.Millisecond
 		testCaseSettings.RetrySleepInitDuration = 10 * time.Millisecond
 		// exponential backoff with 10ms init and 500ms timeout should fit at least 3 attempts.
@@ -241,7 +243,7 @@ func TestDownloadWithRetries(t *testing.T) {
 			return &mockDownloader{"", errors.New("download failed")}, nil
 		}
 
-		u, err := NewUpgrader(testLogger, &settings, &info.AgentInfo{})
+		u, err := NewUpgrader(testLogger, &downloadSettings, &upgradeSettings, &info.AgentInfo{})
 		require.NoError(t, err)
 
 		parsedVersion, err := agtversion.ParseVersion("8.9.0")