Switch the GRPC communication where Agent is running the server and t…

…he beats are connecting back to Agent (elastic#18973)

* Update libbeat fleet management to use elastic-agent-client.

* Work on switching to the new GRPC.

* More on refactor

* Add back the state.

* Format and check

* Add changelog.

* Cleanup.

* mage fmt in x-pack/libbeat.

* Update go.mod.

* Fix from review.

* Fix NewFromConfig comment

* Update go.mod

* Fix imports.

* Fix some locking issues from review.

* Fix lots of issues, add unit testing to cover restarts on failure reporting or crashes.

* Update go.sum

* Fix TestConfigurableRun.

* Fix range over registered apps in GRPC server using RWLock, switch to sync.Map.

* Fix config.

* Fix tests to work on Windows.

* Fix operation start to not start the same application twice.

* Fix enabling and disabling of monitoring.

.gitignore

@@ -11,6 +11,8 @@
 *beat/logs
 *beat/data
 x-pack/functionbeat/pkg
+x-pack/elastic-agent/data
+x-pack/elastic-agent/pkg/agent/operation/tests/downloads
 
 # Files
 .DS_Store
@@ -25,6 +27,9 @@ mage_output_file.go
 x-pack/functionbeat/*/fields.yml
 x-pack/functionbeat/provider/*/functionbeat-*
 x-pack/dockerlogbeat/temproot.tar
+x-pack/elastic-agent/elastic_agent
+x-pack/elastic-agent/fleet.yml
+x-pack/elastic-agent/pkg/agent/operation/tests/scripts/configurable-1.0-darwin-x86_64/configurable
 
 # Editor swap files
 *.swp

libbeat/cmd/instance/beat.go

@@ -462,7 +462,7 @@ func (b *Beat) launch(settings Settings, bt beat.Creator) error {
 	logp.Info("%s start running.", b.Info.Beat)
 
 	// Launch config manager
-	b.ConfigManager.Start()
+	b.ConfigManager.Start(beater.Stop)
 	defer b.ConfigManager.Stop()
 
 	return beater.Run(&b.Beat)

libbeat/management/management.go

@@ -40,7 +40,7 @@ type ConfigManager interface {
 	Enabled() bool
 
 	// Start the config manager
-	Start()
+	Start(func())
 
 	// Stop the config manager
 	Stop()
@@ -98,6 +98,6 @@ func nilFactory(*common.Config, *reload.Registry, uuid.UUID) (ConfigManager, err
 }
 
 func (nilManager) Enabled() bool                           { return false }
-func (nilManager) Start()                                  {}
+func (nilManager) Start(_ func())                          {}
 func (nilManager) Stop()                                   {}
 func (nilManager) CheckRawConfig(cfg *common.Config) error { return nil }

x-pack/elastic-agent/CHANGELOG.asciidoc

@@ -70,3 +70,4 @@
 - Change monitoring defaults for agent {pull}18927[18927]
 - Agent verifies packages before using them {pull}18876[18876]
 - Change stream.* to dataset.* fields {pull}18967[18967]
+- Agent now runs the GRPC server and spawned application connect by to Agent {pull}18973[18973]

x-pack/elastic-agent/_meta/config/common.p2.yml.tmpl

@@ -97,13 +97,18 @@ datasources:
 #   install_path: "${path.data}/install"
 
 # process:
-#   # minimal port number for spawned processes
-#   min_port: 10000
-#   # maximum port number for spawned processes
-#   max_port: 30000
 #   # timeout for creating new processes. when process is not successfully created by this timeout
 #   # start operation is considered a failure
 #   spawn_timeout: 30s
+#   # timeout for stopping processes. when process is not stopped by this timeout then the process.
+#   # is force killed
+#   stop_timeout: 30s
+
+# grpc:
+#   # listen address for the GRPC server that spawned processes connect back to.
+#   address: localhost
+#   # port for the GRPC server that spawned processes connect back to.
+#   port: 6789
 
 # retry:
 #   # Enabled determines whether retry is possible. Default is false.

x-pack/elastic-agent/_meta/config/common.reference.p2.yml.tmpl

@@ -89,13 +89,18 @@ datasources:
 #   install_path: "${path.data}/install"
 
 # process:
-#   # minimal port number for spawned processes
-#   min_port: 10000
-#   # maximum port number for spawned processes
-#   max_port: 30000
 #   # timeout for creating new processes. when process is not successfully created by this timeout
 #   # start operation is considered a failure
 #   spawn_timeout: 30s
+#   # timeout for stopping processes. when process is not stopped by this timeout then the process.
+#   # is force killed
+#   stop_timeout: 30s
+
+# grpc:
+#   # listen address for the GRPC server that spawned processes connect back to.
+#   address: localhost
+#   # port for the GRPC server that spawned processes connect back to.
+#   port: 6789
 
 # retry:
 #   # Enabled determines whether retry is possible. Default is false.

x-pack/elastic-agent/_meta/config/elastic-agent.docker.yml.tmpl

@@ -89,13 +89,18 @@ datasources:
 #   install_path: "${path.data}/install"
 
 # process:
-#   # minimal port number for spawned processes
-#   min_port: 10000
-#   # maximum port number for spawned processes
-#   max_port: 30000
 #   # timeout for creating new processes. when process is not successfully created by this timeout
 #   # start operation is considered a failure
 #   spawn_timeout: 30s
+#   # timeout for stopping processes. when process is not stopped by this timeout then the process.
+#   # is force killed
+#   stop_timeout: 30s
+
+# grpc:
+#   # listen address for the GRPC server that spawned processes connect back to.
+#   address: localhost
+#   # port for the GRPC server that spawned processes connect back to.
+#   port: 6789
 
 # retry:
 #   # Enabled determines whether retry is possible. Default is false.

x-pack/elastic-agent/elastic-agent.docker.yml

@@ -89,13 +89,18 @@ datasources:
 #   install_path: "${path.data}/install"
 
 # process:
-#   # minimal port number for spawned processes
-#   min_port: 10000
-#   # maximum port number for spawned processes
-#   max_port: 30000
 #   # timeout for creating new processes. when process is not successfully created by this timeout
 #   # start operation is considered a failure
 #   spawn_timeout: 30s
+#   # timeout for stopping processes. when process is not stopped by this timeout then the process.
+#   # is force killed
+#   stop_timeout: 30s
+
+# grpc:
+#   # listen address for the GRPC server that spawned processes connect back to.
+#   address: localhost
+#   # port for the GRPC server that spawned processes connect back to.
+#   port: 6789
 
 # retry:
 #   # Enabled determines whether retry is possible. Default is false.

x-pack/elastic-agent/elastic-agent.reference.yml

@@ -95,13 +95,18 @@ datasources:
 #   install_path: "${path.data}/install"
 
 # process:
-#   # minimal port number for spawned processes
-#   min_port: 10000
-#   # maximum port number for spawned processes
-#   max_port: 30000
 #   # timeout for creating new processes. when process is not successfully created by this timeout
 #   # start operation is considered a failure
 #   spawn_timeout: 30s
+#   # timeout for stopping processes. when process is not stopped by this timeout then the process.
+#   # is force killed
+#   stop_timeout: 30s
+
+# grpc:
+#   # listen address for the GRPC server that spawned processes connect back to.
+#   address: localhost
+#   # port for the GRPC server that spawned processes connect back to.
+#   port: 6789
 
 # retry:
 #   # Enabled determines whether retry is possible. Default is false.

x-pack/elastic-agent/elastic-agent.yml

@@ -103,13 +103,18 @@ datasources:
 #   install_path: "${path.data}/install"
 
 # process:
-#   # minimal port number for spawned processes
-#   min_port: 10000
-#   # maximum port number for spawned processes
-#   max_port: 30000
 #   # timeout for creating new processes. when process is not successfully created by this timeout
 #   # start operation is considered a failure
 #   spawn_timeout: 30s
+#   # timeout for stopping processes. when process is not stopped by this timeout then the process.
+#   # is force killed
+#   stop_timeout: 30s
+
+# grpc:
+#   # listen address for the GRPC server that spawned processes connect back to.
+#   address: localhost
+#   # port for the GRPC server that spawned processes connect back to.
+#   port: 6789
 
 # retry:
 #   # Enabled determines whether retry is possible. Default is false.

x-pack/elastic-agent/magefile.go

@@ -12,6 +12,7 @@ import (
 	"os"
 	"os/exec"
 	"path/filepath"
+	"runtime"
 	"strings"
 	"time"
 
@@ -165,8 +166,12 @@ func (Build) Clean() {
 func (Build) TestBinaries() error {
 	p := filepath.Join("pkg", "agent", "operation", "tests", "scripts")
 
+	binaryName := "configurable"
+	if runtime.GOOS == "windows" {
+		binaryName += ".exe"
+	}
 	return combineErr(
-		RunGo("build", "-o", filepath.Join(p, "configurable-1.0-darwin-x86", "configurable"), filepath.Join(p, "configurable-1.0-darwin-x86", "main.go")),
+		RunGo("build", "-o", filepath.Join(p, "configurable-1.0-darwin-x86_64", binaryName), filepath.Join(p, "configurable-1.0-darwin-x86_64", "main.go")),
 	)
 }
 

x-pack/elastic-agent/pkg/agent/application/local_mode.go

@@ -13,7 +13,9 @@ import (
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/agent/errors"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/config"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/core/logger"
+	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/core/plugin/app"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/core/plugin/app/monitoring"
+	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/core/server"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/dir"
 	reporting "github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/reporter"
 	logreporter "github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/reporter/log"
@@ -39,6 +41,7 @@ type Local struct {
 	log         *logger.Logger
 	source      source
 	agentInfo   *info.AgentInfo
+	srv         *server.Server
 }
 
 type source interface {
@@ -78,6 +81,10 @@ func newLocal(
 	}
 
 	localApplication.bgContext, localApplication.cancelCtxFn = context.WithCancel(ctx)
+	localApplication.srv, err = server.NewFromConfig(log, rawConfig, &app.ApplicationStatusHandler{})
+	if err != nil {
+		return nil, errors.New(err, "initialize GRPC listener")
+	}
 
 	reporter := reporting.NewReporter(localApplication.bgContext, log, localApplication.agentInfo, logR)
 
@@ -86,7 +93,7 @@ func newLocal(
 		return nil, errors.New(err, "failed to initialize monitoring")
 	}
 
-	router, err := newRouter(log, streamFactory(localApplication.bgContext, rawConfig, nil, reporter, monitor))
+	router, err := newRouter(log, streamFactory(localApplication.bgContext, rawConfig, localApplication.srv, reporter, monitor))
 	if err != nil {
 		return nil, errors.New(err, "fail to initialize pipeline router")
 	}
@@ -113,6 +120,9 @@ func (l *Local) Start() error {
 	l.log.Info("Agent is starting")
 	defer l.log.Info("Agent is stopped")
 
+	if err := l.srv.Start(); err != nil {
+		return err
+	}
 	if err := l.source.Start(); err != nil {
 		return err
 	}
@@ -123,6 +133,7 @@ func (l *Local) Start() error {
 // Stop stops a local agent.
 func (l *Local) Stop() error {
 	l.cancelCtxFn()
+	l.srv.Stop()
 	return l.source.Stop()
 }
 

x-pack/elastic-agent/pkg/agent/application/managed_mode.go

@@ -11,13 +11,16 @@ import (
 	"net/http"
 	"net/url"
 
+	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/core/plugin/app"
+
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/agent/application/filters"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/agent/application/info"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/agent/errors"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/agent/storage"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/config"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/core/logger"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/core/plugin/app/monitoring"
+	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/core/server"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/fleetapi"
 	reporting "github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/reporter"
 	fleetreporter "github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/reporter/fleet"
@@ -44,6 +47,7 @@ type Managed struct {
 	api         apiClient
 	agentInfo   *info.AgentInfo
 	gateway     *fleetGateway
+	srv         *server.Server
 }
 
 func newManaged(
@@ -110,6 +114,10 @@ func newManaged(
 	}
 
 	managedApplication.bgContext, managedApplication.cancelCtxFn = context.WithCancel(ctx)
+	managedApplication.srv, err = server.NewFromConfig(log, rawConfig, &app.ApplicationStatusHandler{})
+	if err != nil {
+		return nil, errors.New(err, "initialize GRPC listener")
+	}
 
 	logR := logreporter.NewReporter(log, cfg.Reporting.Log)
 	fleetR, err := fleetreporter.NewReporter(agentInfo, log, cfg.Reporting.Fleet)
@@ -123,7 +131,7 @@ func newManaged(
 		return nil, errors.New(err, "failed to initialize monitoring")
 	}
 
-	router, err := newRouter(log, streamFactory(managedApplication.bgContext, rawConfig, client, combinedReporter, monitor))
+	router, err := newRouter(log, streamFactory(managedApplication.bgContext, rawConfig, managedApplication.srv, combinedReporter, monitor))
 	if err != nil {
 		return nil, errors.New(err, "fail to initialize pipeline router")
 	}
@@ -200,6 +208,9 @@ func newManaged(
 // Start starts a managed elastic-agent.
 func (m *Managed) Start() error {
 	m.log.Info("Agent is starting")
+	if err := m.srv.Start(); err != nil {
+		return err
+	}
 	m.gateway.Start()
 	return nil
 }
@@ -208,6 +219,7 @@ func (m *Managed) Start() error {
 func (m *Managed) Stop() error {
 	defer m.log.Info("Agent is stopped")
 	m.cancelCtxFn()
+	m.srv.Stop()
 	return nil
 }
 

x-pack/elastic-agent/pkg/agent/application/stream.go

@@ -19,6 +19,7 @@ import (
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/config"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/core/logger"
 	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/core/plugin/app/monitoring"
+	"github.com/elastic/beats/v7/x-pack/elastic-agent/pkg/core/server"
 )
 
 // EventProcessor is an processor of application event
@@ -56,10 +57,10 @@ func (b *operatorStream) Execute(cfg *configRequest) error {
 	return b.configHandler.HandleConfig(cfg)
 }
 
-func streamFactory(ctx context.Context, cfg *config.Config, client sender, r reporter, m monitoring.Monitor) func(*logger.Logger, routingKey) (stream, error) {
+func streamFactory(ctx context.Context, cfg *config.Config, srv *server.Server, r reporter, m monitoring.Monitor) func(*logger.Logger, routingKey) (stream, error) {
 	return func(log *logger.Logger, id routingKey) (stream, error) {
 		// new operator per stream to isolate processes without using tags
-		operator, err := newOperator(ctx, log, id, cfg, r, m)
+		operator, err := newOperator(ctx, log, id, cfg, srv, r, m)
 		if err != nil {
 			return nil, err
 		}
@@ -71,7 +72,7 @@ func streamFactory(ctx context.Context, cfg *config.Config, client sender, r rep
 	}
 }
 
-func newOperator(ctx context.Context, log *logger.Logger, id routingKey, config *config.Config, r reporter, m monitoring.Monitor) (*operation.Operator, error) {
+func newOperator(ctx context.Context, log *logger.Logger, id routingKey, config *config.Config, srv *server.Server, r reporter, m monitoring.Monitor) (*operation.Operator, error) {
 	operatorConfig := operatorCfg.DefaultConfig()
 	if err := config.Unpack(&operatorConfig); err != nil {
 		return nil, err
@@ -102,6 +103,7 @@ func newOperator(ctx context.Context, log *logger.Logger, id routingKey, config
 		verifier,
 		installer,
 		stateResolver,
+		srv,
 		r,
 		m,
 	)