Review service container aliases and hostname (#1850)

Use specific hostnames for service containers adding alias
networks when connecting the container to the docker
network.

internal/docker/docker.go

@@ -126,14 +126,16 @@ func InspectNetwork(network string) ([]NetworkDescription, error) {
 
 // ConnectToNetwork function connects the container to the selected Docker network.
 func ConnectToNetwork(containerID, network string) error {
-	return ConnectToNetworkWithAlias(containerID, network, "")
+	return ConnectToNetworkWithAlias(containerID, network, []string{})
 }
 
 // ConnectToNetworkWithAlias function connects the container to the selected Docker network.
-func ConnectToNetworkWithAlias(containerID, network, alias string) error {
+func ConnectToNetworkWithAlias(containerID, network string, aliases []string) error {
 	args := []string{"network", "connect", network, containerID}
-	if alias != "" {
-		args = append(args, "--alias", alias)
+	if len(aliases) > 0 {
+		for _, alias := range aliases {
+			args = append(args, "--alias", alias)
+		}
 	}
 	cmd := exec.Command("docker", args...)
 	errOutput := new(bytes.Buffer)

internal/servicedeployer/compose.go

@@ -136,26 +136,37 @@ func (d *DockerComposeServiceDeployer) SetUp(ctx context.Context, svcInfo Servic
 		return nil, fmt.Errorf("service is unhealthy: %w", err)
 	}
 
+	// Added a specific alias when connecting the service to the network.
+	// - There could be container names too long that could not be resolved by the local DNS
+	// - Not used serviceName directly as alias container, since there could be packages defining
+	//   kibana or elasticsearch services and those DNS names are already present in the Elastic stack.
+	//   This is mainly applicable when the Elastic Agent of the stack is used for testing.
+	// - Keep the same alias for both implementations for consistency
+	aliasContainer := fmt.Sprintf("svc-%s", serviceName)
 	if d.runTearDown || d.runTestsOnly {
 		logger.Debug("Skipping connect container to network (non setup steps)")
 	} else {
+		aliases := []string{
+			aliasContainer,
+		}
 		if d.deployIndependentAgent {
 			// Connect service network with agent network
-			err = docker.ConnectToNetworkWithAlias(p.ContainerName(serviceName), svcInfo.AgentNetworkName, svcInfo.Name)
+			err = docker.ConnectToNetworkWithAlias(p.ContainerName(serviceName), svcInfo.AgentNetworkName, aliases)
+
 			if err != nil {
 				return nil, fmt.Errorf("can't attach service container to the agent network: %w", err)
 			}
 		} else {
 			// Connect service network with stack network (for the purpose of metrics collection)
-			err = docker.ConnectToNetwork(p.ContainerName(serviceName), stack.Network(d.profile))
+			err = docker.ConnectToNetworkWithAlias(p.ContainerName(serviceName), stack.Network(d.profile), aliases)
 			if err != nil {
 				return nil, fmt.Errorf("can't attach service container to the stack network: %w", err)
 			}
 		}
 	}
 
 	// Build service container name
-	svcInfo.Hostname = p.ContainerName(serviceName)
+	svcInfo.Hostname = aliasContainer
 
 	logger.Debugf("adding service container %s internal ports to context", p.ContainerName(serviceName))
 	serviceComposeConfig, err := p.Config(ctx, compose.CommandOptions{

internal/testrunner/runners/system/runner.go

@@ -1962,8 +1962,7 @@ func filterAgents(allAgents []kibana.Agent, svcInfo servicedeployer.ServiceInfo)
 			continue // For some reason Kibana doesn't always return a valid policy revision (eventually it will be present and valid)
 		}
 
-		hasServicePrefix := strings.HasPrefix(agent.LocalMetadata.Host.Name, svcInfo.Agent.Host.NamePrefix)
-		if svcInfo.Agent.Host.NamePrefix != "" && !hasServicePrefix {
+		if svcInfo.Agent.Host.NamePrefix != "" && !strings.HasPrefix(agent.LocalMetadata.Host.Name, svcInfo.Agent.Host.NamePrefix) {
 			continue
 		}
 		filtered = append(filtered, agent)

scripts/test-system-test-flags.sh

@@ -2,8 +2,6 @@
 
 set -euxo pipefail
 
-DEFAULT_AGENT_CONTAINER_NAME="elastic-package-service-[0-9]{5}-docker-custom-agent"
-
 cleanup() {
     local r=$?
     local container_id=""
@@ -270,6 +268,14 @@ run_tests_for_package() {
 }
 
 
+# Set variables depending on whether or not independent Elastic Agents are running
+DEFAULT_AGENT_CONTAINER_NAME="elastic-package-service-[0-9]{5}-docker-custom-agent"
+service_deployer_type="docker"
+service_prefix='elastic-package-service-[0-9]{5}'
+if [[ "${ELASTIC_PACKAGE_TEST_ENABLE_INDEPENDENT_AGENT:-"false"}" == "true" ]]; then
+    service_deployer_type="agent"
+fi
+
 # to be set the specific value in run_tests_for_package , required to be global
 # so cleanup function could delete the container if is still running
 SERVICE_CONTAINER_NAME=""
@@ -285,18 +291,10 @@ elastic-package stack status
 
 FOLDER_PATH="${HOME}/.elastic-package/profiles/default/stack/state"
 
-# Check also if independent Elastic Agents are running too
-# depending on the environment variable
-service_deployer_type="docker"
-service_preffix='elastic-package-service-[0-9]{5}'
-if [[ "${ELASTIC_PACKAGE_TEST_ENABLE_INDEPENDENT_AGENT:-"false"}" == "true" ]]; then
-    service_deployer_type="agent"
-fi
-
 # docker service deployer
 if ! run_tests_for_package \
     "test/packages/parallel/nginx" \
-    "${service_preffix}-nginx" \
+    "${service_prefix}-nginx" \
     "data_stream/access/_dev/test/system/test-default-config.yml" \
     "no variant" \
     "${service_deployer_type}" ; then
@@ -306,7 +304,7 @@ fi
 
 if ! run_tests_for_package \
     "test/packages/parallel/sql_input" \
-    "${service_preffix}-sql_input" \
+    "${service_prefix}-sql_input" \
     "_dev/test/system/test-default-config.yml" \
     "mysql_8_0_13" \
     "${service_deployer_type}" ; then