From 5e409274d4de407a7961b98fb0fe1c0465acbe7c Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=EC=A0=84=EC=9B=85=EC=9E=AC?= <juser0719@lguplus.co.kr>
Date: Tue, 7 Jan 2025 17:02:34 +0900
Subject: [PATCH] DTC-8156 add ansible monitor function

---
 pkg/aws/ssm.go           | 76 ++++++++++++++++++++++++++++++++++++++++
 pkg/deployer/deployer.go | 15 ++++++++
 2 files changed, 91 insertions(+)

diff --git a/pkg/aws/ssm.go b/pkg/aws/ssm.go
index 5f3f15c..a92a247 100644
--- a/pkg/aws/ssm.go
+++ b/pkg/aws/ssm.go
@@ -17,10 +17,14 @@ limitations under the license.
 package aws
 
 import (
+	"fmt"
 	"github.com/aws/aws-sdk-go/aws"
 	"github.com/aws/aws-sdk-go/aws/client"
 	"github.com/aws/aws-sdk-go/aws/credentials"
 	"github.com/aws/aws-sdk-go/service/ssm"
+	"github.com/sirupsen/logrus"
+	"strings"
+	"time"
 )
 
 type SSMClient struct {
@@ -58,3 +62,75 @@ func (s SSMClient) SendCommand(target []*string, commands []*string) bool {
 
 	return true
 }
+
+func (s SSMClient) MonitorAnsibleLog(instanceId string) (bool, error) {
+	input := &ssm.SendCommandInput{
+		DocumentName: aws.String("AWS-RunShellScript"),
+		InstanceIds:  []*string{aws.String(instanceId)},
+		Parameters: map[string][]*string{
+			"commands": []*string{
+				aws.String("tail -f /var/log/ansible.log"),
+			},
+		},
+	}
+
+	output, err := s.Client.SendCommand(input)
+	if err != nil {
+		return false, err
+	}
+
+	var lastContent string
+	for {
+		commandOutput, err := s.Client.GetCommandInvocation(&ssm.GetCommandInvocationInput{
+			CommandId:  output.Command.CommandId,
+			InstanceId: aws.String(instanceId),
+		})
+		if err != nil {
+			return false, err
+		}
+
+		if *commandOutput.Status != "Success" {
+			continue
+		}
+
+		currentContent := *commandOutput.StandardOutputContent
+		if newContent := strings.TrimPrefix(currentContent, lastContent); newContent != "" {
+			logrus.Info(fmt.Sprintf("\n%s", newContent))
+			lastContent = currentContent
+		}
+
+		if !strings.Contains(currentContent, "PLAY RECAP") {
+			time.Sleep(time.Second)
+			continue
+		}
+
+		playRecapLines := extractPlayRecap(currentContent)
+		logrus.Info(fmt.Sprintf("Ansible Execution Completed:\n%s", playRecapLines))
+
+		if !strings.Contains(playRecapLines, "failed=0") || !strings.Contains(playRecapLines, "unreachable=0") {
+			return false, fmt.Errorf("ansible execution failed: %s", playRecapLines)
+		}
+
+		return true, nil
+	}
+}
+
+func extractPlayRecap(log string) string {
+	lines := strings.Split(log, "\n")
+	var recap []string
+	inRecap := false
+
+	for _, line := range lines {
+		if strings.Contains(line, "PLAY RECAP") {
+			inRecap = true
+		}
+		if inRecap {
+			recap = append(recap, line)
+			if strings.TrimSpace(line) == "" {
+				break
+			}
+		}
+	}
+
+	return strings.Join(recap, "\n")
+}
diff --git a/pkg/deployer/deployer.go b/pkg/deployer/deployer.go
index 9b8b4c2..493cae5 100644
--- a/pkg/deployer/deployer.go
+++ b/pkg/deployer/deployer.go
@@ -958,6 +958,21 @@ func (d *Deployer) HealthChecking(config schemas.Config) (bool, error) {
 		}
 		d.Logger.Debugf("Health check target autoscaling group: %s / %s", region.Region, *asg.AutoScalingGroupName)
 
+		if len(asg.Instances) > 0 {
+			for _, instance := range asg.Instances {
+				success, err := client.SSMService.MonitorAnsibleLog(*instance.InstanceId)
+				if err != nil {
+					d.Logger.Error(fmt.Sprintf("Error monitoring ansible log: %v", err))
+					return false, err
+				}
+				if !success {
+					d.Logger.Error("Ansible execution failed")
+					return false, err
+				}
+			}
+			d.Logger.Info("Ansible execution completed successfully")
+		}
+
 		isHealthy, err := d.Polling(region, asg, client, config.ForceManifestCapacity, isUpdate, config.DownSizingUpdate)
 		if err != nil {
 			return false, err