chore: add APM integration to CI

.ci/.e2e-tests.yaml

@@ -18,6 +18,12 @@ SUITES:
       - name: "Fleet"
         pullRequestFilter: " && ~debian"
         tags: "fleet_mode_agent"
+      - name: "Integrations"
+        pullRequestFilter: " && ~debian"
+        tags: "integrations"
+      - name: "APM Integration"
+        pullRequestFilter: " && ~debian"
+        tags: "apm_server"
       - name: "Endpoint Integration"
         platforms:
           - "ubuntu-18.04"

cli/config/compose/services/elastic-agent/cloud/docker-compose.yml

@@ -21,6 +21,6 @@ services:
     volumes:
       - "${apmVolume}:/apm-legacy"
     ports:
-      - "127.0.0.1:8220:8220"
+      - "127.0.0.1:8221:8220"
       - "127.0.0.1:8200:8200"
       - "127.0.0.1:5066:5066"

cli/config/compose/services/elastic-agent/docker-compose.yml

@@ -10,6 +10,10 @@ services:
     environment:
       - "FLEET_SERVER_ENABLE=${fleetServerMode:-0}"
       - "FLEET_SERVER_INSECURE_HTTP=${fleetServerMode:-0}"
+      - "FLEET_ENROLL=${fleetEnroll:-1}"
+      - "FLEET_ENROLLMENT_TOKEN=${fleetEnrollmentToken:-}"
+      - "FLEET_INSECURE=${fleetInsecure:-0}"
+      - "FLEET_URL=${fleetUrl:-}"
     platform: ${stackPlatform:-linux/amd64}
     ports:
       - "127.0.0.1:${fleetServerPort:-8220}:8220"

e2e/_suites/fleet/features/agent_endpoint_integration.feature

@@ -6,8 +6,7 @@ Scenario Outline: Adding the Endpoint Integration to an Agent makes the host to
   Given a "<os>" agent is deployed to Fleet with "tar" installer
     And the agent is listed in Fleet as "online"
   When the "Endpoint Security" integration is "added" in the policy
-  Then the "Endpoint Security" datasource is shown in the policy as added
-    And the host name is shown in the Administration view in the Security App as "online"
+  Then the host name is shown in the Administration view in the Security App as "online"
 
 @centos
 Examples: Centos

e2e/_suites/fleet/features/apm_integration.feature

@@ -3,28 +3,19 @@ Feature: APM Integration
 Scenarios for APM
 
 @install
-Scenario Outline: Deploying a <image> stand-alone agent with fleet server mode
-  Given a "<image>" stand-alone agent is deployed with fleet server mode
+Scenario Outline: Deploying a <image> stand-alone agent with the Elastic APM integration
+  Given a "<image>" stand-alone agent is deployed
     And the stand-alone agent is listed in Fleet as "online"
-  When the "Elastic APM" integration is added in the policy
-  Then the "Elastic APM" datasource is shown in the policy as added
-    And the "apm-server" process is in the "started" state on the host
-
-
-@default
-Examples: default
-  | image   |
-  | default |
-
-
-
-@cloud
-Scenario Outline: Deploying a <image> stand-alone agent with fleet server mode on cloud
-  When a "<image>" stand-alone agent is deployed with fleet server mode on cloud
+  When the "Elastic APM" integration is "added" in the policy
   Then the "apm-server" process is in the "started" state on the host
 
-
 @default
 Examples: default
   | image   |
   | default |
+
+@ubi8
+@skip:arm64
+Examples: Ubi8
+| image   |
+| ubi8    |

e2e/_suites/fleet/features/integrations.feature

@@ -0,0 +1,13 @@
+@integrations
+Feature: Integrations
+  Scenarios for operating integrations
+
+@install
+Scenario Outline: Adding an Integration to a Policy
+  When the "<integration>" integration is "added" in the policy
+  Then the "<integration>" datasource is shown in the policy as added
+Examples:
+  | integration |
+  | Elastic APM |
+  | Endpoint    |
+  | Linux       |

e2e/_suites/fleet/features/linux_integration.feature

@@ -7,8 +7,7 @@ Scenario Outline: Adding the Linux Integration to an Agent ...
   Given a "<os>" agent is deployed to Fleet with "tar" installer
     And the agent is listed in Fleet as "online"
   When the "Linux" integration is "added" in the policy
-  Then the "Linux" datasource is shown in the policy as added
-    And a Linux data stream exists with some data
+  Then a Linux data stream exists with some data
 
 @centos
 Examples: Centos

e2e/_suites/fleet/features/stand_alone_agent.feature

@@ -7,7 +7,7 @@ Feature: Stand-alone Agent
 @start-agent
 Scenario Outline: Starting the <image> agent starts backend processes
   When a "<image>" stand-alone agent is deployed
-  Then there are "1" instances of the "filebeat" process in the "started" state
+  Then there are "2" instances of the "filebeat" process in the "started" state
     And there are "2" instances of the "metricbeat" process in the "started" state
 
 @default

e2e/_suites/fleet/fleet.go

@@ -49,7 +49,6 @@ type FleetTestSuite struct {
 	Integration         kibana.IntegrationPackage // the installed integration
 	Policy              kibana.Policy
 	PolicyUpdatedAt     string // the moment the policy was updated
-	FleetServerPolicy   kibana.Policy
 	Version             string // current elastic-agent version
 	kibanaClient        *kibana.Client
 	deployer            deploy.Deployment
@@ -69,50 +68,52 @@ func (fts *FleetTestSuite) afterScenario() {
 	fts.currentContext = apm.ContextWithSpan(context.Background(), span)
 	defer span.End()
 
-	serviceName := common.ElasticAgentServiceName
-	agentService := deploy.NewServiceRequest(serviceName)
+	if fts.InstallerType != "" {
+		serviceName := common.ElasticAgentServiceName
+		agentService := deploy.NewServiceRequest(serviceName)
 
-	if !fts.StandAlone {
-		agentInstaller, _ := installer.Attach(fts.currentContext, fts.deployer, agentService, fts.InstallerType)
+		if !fts.StandAlone {
+			agentInstaller, _ := installer.Attach(fts.currentContext, fts.deployer, agentService, fts.InstallerType)
 
-		if log.IsLevelEnabled(log.DebugLevel) {
-			err := agentInstaller.Logs()
-			if err != nil {
-				log.WithField("error", err).Warn("Could not get agent logs in the container")
+			if log.IsLevelEnabled(log.DebugLevel) {
+				err := agentInstaller.Logs()
+				if err != nil {
+					log.WithField("error", err).Warn("Could not get agent logs in the container")
+				}
 			}
-		}
-		// only call it when the elastic-agent is present
-		if !fts.ElasticAgentStopped {
-			err := agentInstaller.Uninstall(fts.currentContext)
-			if err != nil {
-				log.Warnf("Could not uninstall the agent after the scenario: %v", err)
+			// only call it when the elastic-agent is present
+			if !fts.ElasticAgentStopped {
+				err := agentInstaller.Uninstall(fts.currentContext)
+				if err != nil {
+					log.Warnf("Could not uninstall the agent after the scenario: %v", err)
+				}
 			}
+		} else if log.IsLevelEnabled(log.DebugLevel) {
+			_ = fts.deployer.Logs(agentService)
 		}
-	} else if log.IsLevelEnabled(log.DebugLevel) {
-		_ = fts.deployer.Logs(agentService)
-	}
 
-	err := fts.unenrollHostname()
-	if err != nil {
-		manifest, _ := fts.deployer.Inspect(fts.currentContext, agentService)
-		log.WithFields(log.Fields{
-			"err":      err,
-			"hostname": manifest.Hostname,
-		}).Warn("The agentIDs for the hostname could not be unenrolled")
-	}
+		err := fts.unenrollHostname()
+		if err != nil {
+			manifest, _ := fts.deployer.Inspect(fts.currentContext, agentService)
+			log.WithFields(log.Fields{
+				"err":      err,
+				"hostname": manifest.Hostname,
+			}).Warn("The agentIDs for the hostname could not be unenrolled")
+		}
 
-	if !common.DeveloperMode {
-		_ = fts.deployer.Remove(
-			common.FleetProfileServiceRequest,
-			[]deploy.ServiceRequest{
-				deploy.NewServiceRequest(serviceName),
-			},
-			common.ProfileEnv)
-	} else {
-		log.WithField("service", serviceName).Info("Because we are running in development mode, the service won't be stopped")
+		if !common.DeveloperMode {
+			_ = fts.deployer.Remove(
+				common.FleetProfileServiceRequest,
+				[]deploy.ServiceRequest{
+					deploy.NewServiceRequest(serviceName),
+				},
+				common.ProfileEnv)
+		} else {
+			log.WithField("service", serviceName).Info("Because we are running in development mode, the service won't be stopped")
+		}
 	}
 
-	err = fts.kibanaClient.DeleteEnrollmentAPIKey(fts.currentContext, fts.CurrentTokenID)
+	err := fts.kibanaClient.DeleteEnrollmentAPIKey(fts.currentContext, fts.CurrentTokenID)
 	if err != nil {
 		log.WithFields(log.Fields{
 			"err":     err,
@@ -125,6 +126,7 @@ func (fts *FleetTestSuite) afterScenario() {
 	// clean up fields
 	fts.CurrentTokenID = ""
 	fts.CurrentToken = ""
+	fts.InstallerType = ""
 	fts.Image = ""
 	fts.StandAlone = false
 	fts.BeatsProcess = ""
@@ -183,15 +185,23 @@ func (fts *FleetTestSuite) contributeSteps(s *godog.ScenarioContext) {
 	// stand-alone only steps
 	s.Step(`^a "([^"]*)" stand-alone agent is deployed$`, fts.aStandaloneAgentIsDeployed)
 	s.Step(`^a "([^"]*)" stand-alone agent is deployed with fleet server mode$`, fts.bootstrapFleetServerFromAStandaloneAgent)
-	s.Step(`^a "([^"]*)" stand-alone agent is deployed with fleet server mode on cloud$`, fts.aStandaloneAgentIsDeployedWithFleetServerModeOnCloud)
 	s.Step(`^there is new data in the index from agent$`, fts.thereIsNewDataInTheIndexFromAgent)
 	s.Step(`^the "([^"]*)" docker container is stopped$`, fts.theDockerContainerIsStopped)
 	s.Step(`^there is no new data in the index after agent shuts down$`, fts.thereIsNoNewDataInTheIndexAfterAgentShutsDown)
 	s.Step(`^the stand-alone agent is listed in Fleet as "([^"]*)"$`, fts.theStandaloneAgentIsListedInFleetWithStatus)
 }
 
 func (fts *FleetTestSuite) theStandaloneAgentIsListedInFleetWithStatus(desiredStatus string) error {
+	maxTimeout := time.Duration(utils.TimeoutFactor) * time.Minute
+	exp := utils.GetExponentialBackOff(maxTimeout)
+	retryCount := 0
+
+	agentService := deploy.NewServiceRequest(common.ElasticAgentServiceName)
+	manifest, _ := fts.deployer.Inspect(fts.currentContext, agentService)
+
 	waitForAgents := func() error {
+		retryCount++
+
 		agents, err := fts.kibanaClient.ListAgents(fts.currentContext)
 		if err != nil {
 			return err
@@ -201,13 +211,23 @@ func (fts *FleetTestSuite) theStandaloneAgentIsListedInFleetWithStatus(desiredSt
 			return errors.New("No agents found")
 		}
 
-		agentZero := agents[0]
-		hostname := agentZero.LocalMetadata.Host.HostName
+		for _, agent := range agents {
+			hostname := agent.LocalMetadata.Host.HostName
+
+			if hostname == manifest.Hostname {
+				return theAgentIsListedInFleetWithStatus(fts.currentContext, desiredStatus, hostname)
+			}
+		}
+
+		err = errors.New("Agent not found in Fleet")
+		log.WithFields(log.Fields{
+			"elapsedTime": exp.GetElapsedTime(),
+			"hostname":    manifest.Hostname,
+			"retries":     retryCount,
+		}).Warn(err)
 
-		return theAgentIsListedInFleetWithStatus(fts.currentContext, desiredStatus, hostname)
+		return err
 	}
-	maxTimeout := time.Duration(utils.TimeoutFactor) * time.Minute * 2
-	exp := utils.GetExponentialBackOff(maxTimeout)
 
 	err := backoff.Retry(waitForAgents, exp)
 	if err != nil {