Arbll/collect journalctl logs

pkg/fleet/installer/packages/datadog_agent_linux.go

@@ -10,6 +10,7 @@ import (
 	"fmt"
 	"os"
 	"path/filepath"
+	"slices"
 	"strings"
 
 	"github.com/DataDog/datadog-agent/pkg/fleet/installer/installinfo"
@@ -481,11 +482,11 @@ func (s *datadogAgentService) StopStable(ctx HookContext) error {
 	}
 	switch service.GetServiceManagerType() {
 	case service.SystemdType:
-		return systemd.StopUnits(ctx, s.SystemdUnitsStable...)
+		return systemd.StopUnits(ctx, reverseStringSlice(s.SystemdUnitsStable)...)
 	case service.UpstartType:
-		return upstart.StopAll(ctx, s.UpstartServices...)
+		return upstart.StopAll(ctx, reverseStringSlice(s.UpstartServices)...)
 	case service.SysvinitType:
-		return sysvinit.StopAll(ctx, s.SysvinitServices...)
+		return sysvinit.StopAll(ctx, reverseStringSlice(s.SysvinitServices)...)
 	default:
 		return fmt.Errorf("unsupported service manager")
 	}
@@ -659,3 +660,10 @@ func writeEmbeddedUnit(dir string, unit string, content []byte) error {
 	}
 	return nil
 }
+
+func reverseStringSlice(slice []string) []string {
+	reversed := make([]string, len(slice))
+	copy(reversed, slice)
+	slices.Reverse(reversed)
+	return reversed
+}

pkg/fleet/installer/packages/service/systemd/systemd.go

@@ -16,6 +16,7 @@ import (
 	"os/exec"
 	"path/filepath"
 	"syscall"
+	"time"
 
 	"github.com/DataDog/datadog-agent/pkg/fleet/installer/telemetry"
 	"github.com/DataDog/datadog-agent/pkg/util/log"
@@ -142,3 +143,13 @@ func IsRunning() (running bool, err error) {
 	}
 	return true, nil
 }
+
+// JournaldLogs returns the logs for a given unit since a given time
+func JournaldLogs(ctx context.Context, unit string, since time.Time) (string, error) {
+	journalctlCmd := exec.CommandContext(ctx, "journalctl", "_COMM=systemd", "--unit", unit, "-e", "--no-pager", "--since", since.Format(time.RFC3339))
+	stdout, err := journalctlCmd.Output()
+	if err != nil {
+		return "", err
+	}
+	return string(stdout), nil
+}

pkg/fleet/installer/setup/common/services_nix.go

@@ -8,32 +8,29 @@
 package common
 
 import (
-	"bytes"
 	"fmt"
-	"os/exec"
+	"time"
+
+	"github.com/DataDog/datadog-agent/pkg/fleet/installer/packages/service/systemd"
+	"github.com/DataDog/datadog-agent/pkg/fleet/installer/telemetry"
 )
 
 // restartServices restarts the services that need to be restarted after a package upgrade or
 // an install script re-run; because the configuration may have changed.
 func (s *Setup) restartServices(pkgs []packageWithVersion) error {
+	t := time.Now()
+	span, ctx := telemetry.StartSpanFromContext(s.Ctx, "restartServices")
 	for _, pkg := range pkgs {
 		switch pkg.name {
 		case DatadogAgentPackage:
-			if err := restartService("datadog-agent.service"); err != nil {
-				return err
+			err := systemd.RestartUnit(ctx, "datadog-agent.service")
+			if err != nil {
+				logs, logsErr := systemd.JournaldLogs(ctx, "datadog-agent.service", t)
+				span.SetTag("journald_logs", logs)
+				span.SetTag("journald_logs_err", logsErr)
+				return fmt.Errorf("failed to restart datadog-agent.service: %w", err)
 			}
 		}
 	}
 	return nil
 }
-
-func restartService(unit string) error {
-	cmd := exec.Command("systemctl", "restart", unit)
-	stderr := bytes.Buffer{}
-	cmd.Stderr = &stderr
-	err := cmd.Run()
-	if err != nil {
-		return fmt.Errorf("failed to restart %s (%s): %s", unit, err.Error(), stderr.String())
-	}
-	return nil
-}