Use a builder pattern to initialize resources used by investigations.

To allow not all resources to be prefilled as not every investigation
needs everything, this introduces a Builder pattern that is prefilled
with a subset of fields (OCM, PD, Cluster, Logging) that can be extended
by each investigation.

Once everything that is supposed to be build is set, calling Build()
will return the requested resources.

Author: bergmannf

cadctl/cmd/investigate/investigate.go

@@ -22,12 +22,11 @@ import (
 	"strconv"
 	"strings"
 
-	cmv1 "github.com/openshift-online/ocm-sdk-go/clustersmgmt/v1"
 	investigations "github.com/openshift/configuration-anomaly-detection/pkg/investigations"
 	"github.com/openshift/configuration-anomaly-detection/pkg/investigations/ccam"
 	investigation "github.com/openshift/configuration-anomaly-detection/pkg/investigations/investigation"
+	"github.com/openshift/configuration-anomaly-detection/pkg/investigations/precheck"
 	"github.com/openshift/configuration-anomaly-detection/pkg/logging"
-	"github.com/openshift/configuration-anomaly-detection/pkg/managedcloud"
 	"github.com/openshift/configuration-anomaly-detection/pkg/metrics"
 	ocm "github.com/openshift/configuration-anomaly-detection/pkg/ocm"
 	"github.com/openshift/configuration-anomaly-detection/pkg/pagerduty"
@@ -115,45 +114,25 @@ func run(cmd *cobra.Command, _ []string) error {
 		return fmt.Errorf("could not initialize ocm client: %w", err)
 	}
 
-	cluster, err := ocmClient.GetClusterInfo(clusterID)
-	if err != nil {
-		if strings.Contains(err.Error(), "no cluster found") {
-			logging.Warnf("No cluster found with ID '%s'. Exiting.", clusterID)
-			return pdClient.EscalateIncidentWithNote("CAD was unable to find the incident cluster in OCM. An alert for a non-existing cluster is unexpected. Please investigate manually.")
-		}
-		return fmt.Errorf("could not retrieve cluster info for %s: %w", clusterID, err)
-	}
-
-	// From this point on, we normalize to internal ID, as this ID always exists.
-	// For installing clusters, externalID can be empty.
-	internalClusterID := cluster.ID()
-
-	// re-initialize logger for the internal-cluster-id context
-	logging.RawLogger = logging.InitLogger(logLevelFlag, pipelineNameEnv, internalClusterID)
+	builder := &investigation.ResourceBuilderT{}
+	// Prime the builder with information required for all investigations.
+	builder.WithName(alertInvestigation.Name()).WithCluster(clusterID).WithPagerDutyClient(pdClient).WithOcmClient(ocmClient).WithLogger(logLevelFlag, pipelineNameEnv)
 
-	requiresInvestigation, err := clusterRequiresInvestigation(cluster, pdClient, ocmClient)
-	if err != nil || !requiresInvestigation {
+	precheck := precheck.Investigation{}
+	result, err := precheck.Run(builder)
+	if err != nil || result.StopInvestigations {
 		return err
 	}
 
-	clusterDeployment, err := ocmClient.GetClusterDeployment(internalClusterID)
-	if err != nil {
-		return fmt.Errorf("could not retrieve Cluster Deployment for %s: %w", internalClusterID, err)
-	}
-
-	customerAwsClient, err := managedcloud.CreateCustomerAWSClient(cluster, ocmClient)
+	inv := ccam.Investigation{}
+	result, err = inv.Run(builder)
+	updateMetrics(alertInvestigation.Name(), &result)
 	if err != nil {
-		ccamResources := &investigation.Resources{Name: "ccam", Cluster: cluster, ClusterDeployment: clusterDeployment, AwsClient: customerAwsClient, OcmClient: ocmClient, PdClient: pdClient, Notes: nil, AdditionalResources: map[string]interface{}{"error": err}}
-		inv := ccam.Investigation{}
-		result, err := inv.Run(ccamResources)
-		updateMetrics(alertInvestigation.Name(), &result)
 		return err
 	}
 
-	investigationResources = &investigation.Resources{Name: alertInvestigation.Name(), Cluster: cluster, ClusterDeployment: clusterDeployment, AwsClient: customerAwsClient, OcmClient: ocmClient, PdClient: pdClient, Notes: nil}
-
 	logging.Infof("Starting investigation for %s", alertInvestigation.Name())
-	result, err := alertInvestigation.Run(investigationResources)
+	result, err = alertInvestigation.Run(builder)
 	updateMetrics(alertInvestigation.Name(), &result)
 	if err != nil {
 		return err
@@ -185,32 +164,6 @@ func handleCADFailure(err error, resources *investigation.Resources, pdClient *p
 	}
 }
 
-// Checks pre-requisites for a cluster investigation:
-// - the cluster's state is supported by CAD for an investigation (= not uninstalling)
-// - the cloud provider is supported by CAD (cluster is AWS)
-// Performs according pagerduty actions and returns whether CAD needs to investigate the cluster
-func clusterRequiresInvestigation(cluster *cmv1.Cluster, pdClient *pagerduty.SdkClient, ocmClient *ocm.SdkClient) (bool, error) {
-	if cluster.State() == cmv1.ClusterStateUninstalling {
-		logging.Info("Cluster is uninstalling and requires no investigation. Silencing alert.")
-		return false, pdClient.SilenceIncidentWithNote("CAD: Cluster is already uninstalling, silencing alert.")
-	}
-
-	if cluster.AWS() == nil {
-		logging.Info("Cloud provider unsupported, forwarding to primary.")
-		return false, pdClient.EscalateIncidentWithNote("CAD could not run an automated investigation on this cluster: unsupported cloud provider.")
-	}
-
-	isAccessProtected, err := ocmClient.IsAccessProtected(cluster)
-	if err != nil {
-		logging.Warnf("failed to get access protection status for cluster. %w. Continuing...")
-	}
-	if isAccessProtected {
-		logging.Info("Cluster is access protected. Escalating alert.")
-		return false, pdClient.EscalateIncidentWithNote("CAD is unable to run against access protected clusters. Please investigate.")
-	}
-	return true, nil
-}
-
 func updateMetrics(investigationName string, result *investigation.InvestigationResult) {
 	if result.ServiceLogSent.Performed {
 		metrics.Inc(metrics.ServicelogSent, append([]string{investigationName}, result.ServiceLogSent.Labels...)...)