[Cloud Connect] Use Display Name to register cluster (#47440) (#47529)

This overrides the cluster name submitted to the Cloud Connected
API with the `cluster.metadata.display_name` setting -- if set --
with the proper precendence of transient, then persistent setting.

For any existing registration, this will overwrite the old name
with the new name upon restart.

(cherry picked from commit e991bf0)

Co-authored-by: Chris Earle <[email protected]>

Author: mergify[bot]

changelog/fragments/1762213574-cc-cluster-name.yaml

@@ -0,0 +1,19 @@
+# REQUIRED
+# - bug-fix: fixes a problem in a previous version
+kind: bug-fix
+summary: "[Cloud Connect] Use cluster.metadata.display_name as cluster name if set"
+component: metricbeat
+
+# AUTOMATED
+# OPTIONAL to manually add other PR URLs
+# PR URL: A link the PR that added the changeset.
+# If not present is automatically filled by the tooling finding the PR where this changelog fragment has been added.
+# NOTE: the tooling supports backports, so it's able to fill the original PR number instead of the backport PR number.
+# Please provide it if you are adding a fragment for a different PR.
+# pr: https://github.com/owner/repo/1234
+
+# AUTOMATED
+# OPTIONAL to manually add other issue URLs
+# Issue URL; optional; the GitHub issue related to this changeset (either closes or is part of).
+# If not present is automatically filled by the tooling with the issue linked to the PR number.
+# issue: https://github.com/owner/repo/1234

x-pack/metricbeat/module/autoops_es/metricset/register.go

@@ -18,10 +18,25 @@ import (
 )
 
 const (
-	licensePath   = "/_license"
-	licensePathV7 = "/_license?accept_enterprise"
+	clusterSettingsPath = "/_cluster/settings?filter_path=**.cluster.metadata.display_name"
+	licensePath         = "/_license"
+	licensePathV7       = "/_license?accept_enterprise"
 )
 
+type clusterSettingsResponse struct {
+	Persistent clusterSettingsDisplayName `json:"persistent"`
+	Transient  clusterSettingsDisplayName `json:"transient"`
+}
+
+// Partial structure to extract the cluster display name from the cluster settings API response.
+type clusterSettingsDisplayName struct {
+	Cluster struct {
+		Metadata struct {
+			DisplayName string `json:"display_name"`
+		} `json:"metadata"`
+	} `json:"cluster"`
+}
+
 // License information returned from `GET /_license`, contained within the "license" object.
 type license struct {
 	UID    string `json:"uid"`
@@ -75,20 +90,38 @@ func maybeRegisterCloudConnectedCluster(m *elasticsearch.MetricSet, getClusterIn
 	cloudApiKey, err := getCloudConnectedModeApiKey(m)
 
 	if err != nil {
-		return fmt.Errorf("failed to get Cloud Connected Mode API key: %w", err)
+		return fmt.Errorf("failed to get Cloud Connected API Key: %w", err)
 	} else if cloudApiKey == "" {
 		// if there is no Cloud API Key configured, then there is no request to make
 		return nil
 	}
 
-	m.Logger().Debugf("Attempting to get cluster info for Cloud Connected Mode...")
+	m.Logger().Debugf("Attempting to get cluster info for Cloud Connected...")
 	clusterInfo, err := getClusterInfo(m)
 
 	if err != nil {
 		return fmt.Errorf("failed to load cluster info: %w", err)
 	}
 
-	m.Logger().Debugf("Attempting to fetch license for Cloud Connected Mode...")
+	m.Logger().Debugf("Attempting to fetch cluster settings for Cloud Connected...")
+	clusterSettings, err := utils.FetchAPIData[clusterSettingsResponse](m, clusterSettingsPath)
+
+	if err != nil {
+		// Log the error, but do not fail the registration - we can continue with the existing cluster name
+		// Note: You can restart the agent to re-fetch the name later if needed
+		m.Logger().Warnf("Ignoring failure to fetch cluster settings for Cloud Connected: %v", err)
+		// prefer transient setting over persistent setting
+	} else if clusterSettings.Transient.Cluster.Metadata.DisplayName != "" {
+		m.Logger().Debugf("Using transient cluster display name %s for Cloud Connected in place of cluster name %s", clusterSettings.Transient.Cluster.Metadata.DisplayName, clusterInfo.ClusterName)
+
+		clusterInfo.ClusterName = clusterSettings.Transient.Cluster.Metadata.DisplayName
+	} else if clusterSettings.Persistent.Cluster.Metadata.DisplayName != "" {
+		m.Logger().Debugf("Using persistent cluster display name %s for Cloud Connected in place of cluster name %s", clusterSettings.Persistent.Cluster.Metadata.DisplayName, clusterInfo.ClusterName)
+
+		clusterInfo.ClusterName = clusterSettings.Persistent.Cluster.Metadata.DisplayName
+	}
+
+	m.Logger().Debugf("Attempting to fetch license for Cloud Connected...")
 	versionedLicensePath := licensePath
 
 	// TODO: Drop support for this when we drop support for 7.17
@@ -107,7 +140,7 @@ func maybeRegisterCloudConnectedCluster(m *elasticsearch.MetricSet, getClusterIn
 		return fmt.Errorf("cluster license type is not supported: %s", licenseWrapper.License.Type)
 	}
 
-	m.Logger().Debugf("Successfully fetched license for Cloud Connected Mode: UUID=%s License=%s", clusterInfo.ClusterID, licenseWrapper.License.UID)
+	m.Logger().Debugf("Successfully fetched license for Cloud Connected: UUID=%s License=%s", clusterInfo.ClusterID, licenseWrapper.License.UID)
 
 	return registerCloudConnectedCluster(cloudApiKey, clusterInfo, &licenseWrapper.License, m.Logger())
 }
@@ -128,14 +161,14 @@ func registerCloudConnectedCluster(cloudApiKey string, clusterInfo *utils.Cluste
 	})
 
 	if err != nil {
-		return fmt.Errorf("failed to serialize payload for Cloud Connected Mode: %w", err)
+		return fmt.Errorf("failed to serialize payload for Cloud Connected: %w", err)
 	}
 
 	requestURL := getCloudConnectedModeAPIURL() + "/api/v1/cloud-connected/clusters"
 	req, err := http.NewRequestWithContext(context.Background(), "POST", requestURL, bytes.NewBuffer(jsonData))
 
 	if err != nil {
-		return fmt.Errorf("failed to create HTTP request for Cloud Connected Mode: %w", err)
+		return fmt.Errorf("failed to create HTTP request for Cloud Connected: %w", err)
 	}
 
 	req.Header.Set("Accept", "application/json")
@@ -146,9 +179,9 @@ func registerCloudConnectedCluster(cloudApiKey string, clusterInfo *utils.Cluste
 
 	if err == nil {
 		utils.SetResourceID(data.ID)
-		logger.Infof("Registered cluster %s for Cloud Connected Mode: %s", clusterInfo.ClusterID, data.ID)
+		logger.Infof("Registered cluster %s for Cloud Connected: %s", clusterInfo.ClusterID, data.ID)
 		return nil
 	}
 
-	return fmt.Errorf("failed to register for Cloud Connected Mode: %w", err)
+	return fmt.Errorf("failed to register for Cloud Connected: %w", err)
 }