fix(healthcheck): get discoverable nodes from config-cache

pkg/service/healthcheck/runner.go

@@ -10,7 +10,8 @@ import (
 	"github.com/pkg/errors"
 	"github.com/prometheus/client_golang/prometheus"
 	"github.com/scylladb/go-log"
-	"github.com/scylladb/go-set/strset"
+	"github.com/scylladb/scylla-manager/v3/pkg/service/configcache"
+	"github.com/scylladb/scylla-manager/v3/pkg/util/slice"
 
 	"github.com/scylladb/scylla-manager/v3/pkg/scyllaclient"
 	"github.com/scylladb/scylla-manager/v3/pkg/service"
@@ -45,6 +46,7 @@ func (r Runner) Run(ctx context.Context, clusterID, taskID, runID uuid.UUID, pro
 
 type runner struct {
 	logger       log.Logger
+	configCacher configcache.ConfigCacher
 	scyllaClient scyllaclient.ProviderFunc
 	timeout      time.Duration
 	metrics      *runnerMetrics
@@ -67,31 +69,24 @@ func (r runner) Run(ctx context.Context, clusterID, _, _ uuid.UUID, _ json.RawMe
 	// Enable interactive mode for fast backoff
 	ctx = scyllaclient.Interactive(ctx)
 
-	client, err := r.scyllaClient(ctx, clusterID)
+	nodes, err := r.configCacher.AvailableHosts(ctx, clusterID)
 	if err != nil {
-		return errors.Wrap(err, "get client")
+		return err
 	}
-
-	status, err := client.Status(ctx)
-	if err != nil {
-		return errors.Wrap(err, "status")
-	}
-
-	live := status.Live()
-	r.removeMetricsForMissingHosts(clusterID, live)
-	r.checkHosts(ctx, clusterID, live)
+	r.removeMetricsForMissingHosts(clusterID, nodes)
+	r.checkHosts(ctx, clusterID, nodes)
 
 	return nil
 }
 
-func (r runner) checkHosts(ctx context.Context, clusterID uuid.UUID, status []scyllaclient.NodeStatusInfo) {
+func (r runner) checkHosts(ctx context.Context, clusterID uuid.UUID, addresses []string) {
 	f := func(i int) error {
 		hl := prometheus.Labels{
 			clusterKey: clusterID.String(),
-			hostKey:    status[i].Addr,
+			hostKey:    addresses[i],
 		}
 
-		rtt, err := r.ping(ctx, clusterID, status[i].Addr, r.timeout)
+		rtt, err := r.ping(ctx, clusterID, addresses[i], r.timeout)
 		if err != nil {
 			r.metrics.status.With(hl).Set(-1)
 		} else {
@@ -102,8 +97,8 @@ func (r runner) checkHosts(ctx context.Context, clusterID uuid.UUID, status []sc
 		return nil
 	}
 
-	_ = parallel.Run(len(status), parallel.NoLimit, f, func(i int, err error) { // nolint: errcheck
-		r.logger.Error(ctx, "Parallel hosts check failed", "", status[i].Addr, "error", err)
+	_ = parallel.Run(len(addresses), parallel.NoLimit, f, func(i int, err error) { // nolint: errcheck
+		r.logger.Error(ctx, "Parallel hosts check failed", "", addresses[i], "error", err)
 	})
 }
 
@@ -122,17 +117,12 @@ func (r runner) removeMetricsForCluster(clusterID uuid.UUID) {
 	})
 }
 
-func (r runner) removeMetricsForMissingHosts(clusterID uuid.UUID, status []scyllaclient.NodeStatusInfo) {
-	m := strset.New()
-	for _, node := range status {
-		m.Add(node.Addr)
-	}
-
+func (r runner) removeMetricsForMissingHosts(clusterID uuid.UUID, addresses []string) {
 	apply(collect(r.metrics.status), func(cluster, dc, host, pt string, v float64) {
 		if clusterID.String() != cluster {
 			return
 		}
-		if m.Has(host) {
+		if slice.ContainsString(addresses, host) {
 			return
 		}
 

pkg/service/healthcheck/service.go

@@ -57,6 +57,7 @@ func (s *Service) Runner() Runner {
 	return Runner{
 		cql: runner{
 			logger:       s.logger.Named("CQL healthcheck"),
+			configCacher: s.configCache,
 			scyllaClient: s.scyllaClient,
 			timeout:      s.config.MaxTimeout,
 			metrics: &runnerMetrics{
@@ -68,6 +69,7 @@ func (s *Service) Runner() Runner {
 		},
 		rest: runner{
 			logger:       s.logger.Named("REST healthcheck"),
+			configCacher: s.configCache,
 			scyllaClient: s.scyllaClient,
 			timeout:      s.config.MaxTimeout,
 			metrics: &runnerMetrics{
@@ -79,6 +81,7 @@ func (s *Service) Runner() Runner {
 		},
 		alternator: runner{
 			logger:       s.logger.Named("Alternator healthcheck"),
+			configCacher: s.configCache,
 			scyllaClient: s.scyllaClient,
 			timeout:      s.config.MaxTimeout,
 			metrics: &runnerMetrics{