From cd32fa260b9aec041e320afc3fdde08c7a977645 Mon Sep 17 00:00:00 2001
From: Anderson Queiroz <anderson.queiroz@elastic.co>
Date: Tue, 8 Aug 2023 18:24:11 +0200
Subject: [PATCH] enroll command fails if the agent does not restart after
 enroll

For the agent to be actually enrolled it needs to restart after the enroll process is completed, so it'll pickup the new config and "connect" to fleet-server.

This change makes the enroll command to fail if it cannot restart the agent after enrolling on fleet
---
 internal/pkg/agent/cmd/enroll_cmd.go | 43 ++++++++++++++++++++--------
 internal/pkg/agent/cmd/install.go    |  4 ++-
 2 files changed, 34 insertions(+), 13 deletions(-)

diff --git a/internal/pkg/agent/cmd/enroll_cmd.go b/internal/pkg/agent/cmd/enroll_cmd.go
index 3f9145435c0..297e2952aba 100644
--- a/internal/pkg/agent/cmd/enroll_cmd.go
+++ b/internal/pkg/agent/cmd/enroll_cmd.go
@@ -234,7 +234,7 @@ func (c *enrollCmd) Execute(ctx context.Context, streams *cli.IOStreams) error {
 		// Ensure that the agent does not use a proxy configuration
 		// when connecting to the local fleet server.
 		// Note that when running fleet-server the enroll request will be sent to :8220,
-		// however when the agent is running afterwards requests will be sent to :8221
+		// however when the agent is running afterward requests will be sent to :8221
 		c.remoteConfig.Transport.Proxy.Disable = true
 	}
 
@@ -255,7 +255,7 @@ func (c *enrollCmd) Execute(ctx context.Context, streams *cli.IOStreams) error {
 
 	err = c.enrollWithBackoff(ctx, persistentConfig)
 	if err != nil {
-		return errors.New(err, "fail to enroll")
+		return fmt.Errorf("fail to enroll: %w", err)
 	}
 
 	if c.options.FixPermissions {
@@ -266,17 +266,23 @@ func (c *enrollCmd) Execute(ctx context.Context, streams *cli.IOStreams) error {
 	}
 
 	defer func() {
-		fmt.Fprintln(streams.Out, "Successfully enrolled the Elastic Agent.")
+		if err != nil {
+			fmt.Fprintln(streams.Out, "Successfully enrolled the Elastic Agent.")
+		} else {
+			fmt.Fprintf(streams.Out, "Something went wrong while enrolling the Elastic Agent: %v\n", err)
+		}
 	}()
 
 	if c.agentProc == nil {
-		if c.daemonReload(ctx) != nil {
-			c.log.Info("Elastic Agent might not be running; unable to trigger restart")
-		} else {
-			c.log.Info("Successfully triggered restart on running Elastic Agent.")
+		if err = c.daemonReloadWithBackoff(ctx); err != nil {
+			c.log.Errorf("Elastic Agent might not be running; unable to trigger restart: %v", err)
+			return fmt.Errorf("could not reload agent deamon, unable to trigger restart: %v", err)
 		}
+
+		c.log.Info("Successfully triggered restart on running Elastic Agent.")
 		return nil
 	}
+
 	c.log.Info("Elastic Agent has been enrolled; start Elastic Agent")
 	return nil
 }
@@ -477,8 +483,20 @@ func (c *enrollCmd) enrollWithBackoff(ctx context.Context, persistentConfig map[
 
 	c.log.Infof("Starting enrollment to URL: %s", c.client.URI())
 	err := c.enroll(ctx, persistentConfig)
+	if err == nil {
+		return nil
+	}
+
+	const deadline = 10 * time.Minute
+	const frequency = 60 * time.Second
+
+	c.log.Infof("1st enrollment attempt failed, retrying for %s each %s enrolling to URL: %s",
+		deadline,
+		frequency,
+		c.client.URI())
 	signal := make(chan struct{})
-	backExp := backoff.NewExpBackoff(signal, 60*time.Second, 10*time.Minute)
+	defer close(signal)
+	backExp := backoff.NewExpBackoff(signal, frequency, deadline)
 
 	for {
 		retry := false
@@ -497,7 +515,6 @@ func (c *enrollCmd) enrollWithBackoff(ctx context.Context, persistentConfig map[
 		err = c.enroll(ctx, persistentConfig)
 	}
 
-	close(signal)
 	return err
 }
 
@@ -546,8 +563,10 @@ func (c *enrollCmd) enroll(ctx context.Context, persistentConfig map[string]inte
 			c.options.FleetServer.ElasticsearchInsecure,
 		)
 		if err != nil {
-			return err
+			return fmt.Errorf(
+				"failed creating fleet-server bootstrap config: %w", err)
 		}
+
 		// no longer need bootstrap at this point
 		serverConfig.Server.Bootstrap = false
 		fleetConfig.Server = serverConfig.Server
@@ -567,11 +586,11 @@ func (c *enrollCmd) enroll(ctx context.Context, persistentConfig map[string]inte
 
 	reader, err := yamlToReader(configToStore)
 	if err != nil {
-		return err
+		return fmt.Errorf("yamlToReader failed: %w", err)
 	}
 
 	if err := safelyStoreAgentInfo(c.configStore, reader); err != nil {
-		return err
+		return fmt.Errorf("failed to store agent config: %w", err)
 	}
 
 	// clear action store
diff --git a/internal/pkg/agent/cmd/install.go b/internal/pkg/agent/cmd/install.go
index 90fd9eba10d..5bc000843ae 100644
--- a/internal/pkg/agent/cmd/install.go
+++ b/internal/pkg/agent/cmd/install.go
@@ -83,7 +83,7 @@ func installCmd(streams *cli.IOStreams, cmd *cobra.Command) error {
 
 	nonInteractive, _ := cmd.Flags().GetBool("non-interactive")
 	if nonInteractive {
-		fmt.Fprintf(streams.Out, "Installing in non-interactive mode.")
+		fmt.Fprintf(streams.Out, "Installing in non-interactive mode.\n")
 	}
 
 	if status == install.PackageInstall {
@@ -205,6 +205,8 @@ func installCmd(streams *cli.IOStreams, cmd *cobra.Command) error {
 				}
 			}()
 		}
+
+		fmt.Fprint(streams.Out, "Elastic Agent successfully installed, starting enrollment.\n")
 	}
 
 	if enroll {