satellite/{audit,overlay,satellitedb}: enable reporting offline audits

- Remove flag for switching off offline audit reporting.
- Change the overlay method used from UpdateUptime to BatchUpdateStats, as this
is where the new online scoring is done.
- Add a new overlay.AuditOutcome type: AuditOffline. Since we now use the same
method to record offline audits as success, failure, and unknown, we need to
distinguish offline audits from the rest.

Change-Id: Iadcfe10cf13466fa1a1c2dc542db8994a6423355

satellite/audit/reporter.go

@@ -13,9 +13,6 @@ import (
 	"storj.io/storj/satellite/overlay"
 )
 
-// We do not report offline nodes to the overlay at this time; see V3-3025.
-const reportOfflineDuringAudit = false
-
 // Reporter records audit reports in overlay and implements the reporter interface
 //
 // architecture: Service
@@ -94,8 +91,7 @@ func (reporter *Reporter) RecordAudits(ctx context.Context, req Report, path sto
 				errlist.Add(err)
 			}
 		}
-		// We do not report offline nodes to the overlay at this time; see V3-3025.
-		if len(offlines) > 0 && reportOfflineDuringAudit {
+		if len(offlines) > 0 {
 			offlines, err = reporter.recordOfflineStatus(ctx, offlines)
 			if err != nil {
 				errlist.Add(err)
@@ -165,22 +161,23 @@ func (reporter *Reporter) recordAuditUnknownStatus(ctx context.Context, unknownA
 	return nil, nil
 }
 
-// recordOfflineStatus updates nodeIDs in overlay with isup=false. When there
-// is any error the function return the list of nodes which haven't been
-// recorded.
+// recordOfflineStatus updates nodeIDs in overlay with isup=false, auditoutcome=offline.
 func (reporter *Reporter) recordOfflineStatus(ctx context.Context, offlineNodeIDs storj.NodeIDList) (failed storj.NodeIDList, err error) {
 	defer mon.Task()(&ctx)(&err)
-	var errlist errs.Group
-	for _, nodeID := range offlineNodeIDs {
-		_, err := reporter.overlay.UpdateUptime(ctx, nodeID, false)
-		if err != nil {
-			failed = append(failed, nodeID)
-			errlist.Add(err)
+
+	updateRequests := make([]*overlay.UpdateRequest, len(offlineNodeIDs))
+	for i, nodeID := range offlineNodeIDs {
+		updateRequests[i] = &overlay.UpdateRequest{
+			NodeID:       nodeID,
+			IsUp:         false,
+			AuditOutcome: overlay.AuditOffline,
 		}
 	}
-	if len(failed) > 0 {
+
+	failed, err = reporter.overlay.BatchUpdateStats(ctx, updateRequests)
+	if err != nil || len(failed) > 0 {
 		reporter.log.Debug("failed to record Offline Nodes ", zap.Strings("NodeIDs", failed.Strings()))
-		return failed, errs.Combine(Error.New("failed to record some audit offline statuses in overlay"), errlist.Err())
+		return failed, errs.Combine(Error.New("failed to record some audit offline statuses in overlay"), err)
 	}
 
 	return nil, nil

satellite/audit/reporter_test.go

@@ -242,3 +242,29 @@ func TestGracefullyExitedNotUpdated(t *testing.T) {
 		}
 	})
 }
+
+func TestReportOfflineAudits(t *testing.T) {
+	testplanet.Run(t, testplanet.Config{
+		SatelliteCount: 1, StorageNodeCount: 1, UplinkCount: 0,
+	}, func(t *testing.T, ctx *testcontext.Context, planet *testplanet.Planet) {
+		satellite := planet.Satellites[0]
+		node := planet.StorageNodes[0]
+		audits := satellite.Audit
+		audits.Worker.Loop.Pause()
+		cache := satellite.Overlay.DB
+
+		_, err := audits.Reporter.RecordAudits(ctx, audit.Report{Offlines: storj.NodeIDList{node.ID()}}, "")
+		require.NoError(t, err)
+
+		d, err := cache.Get(ctx, node.ID())
+		require.NoError(t, err)
+		require.Equal(t, int64(1), d.Reputation.AuditCount)
+
+		// check that other reputation stats were not incorrectly updated by offline audit
+		require.EqualValues(t, 0, d.Reputation.AuditSuccessCount)
+		require.EqualValues(t, 1, d.Reputation.AuditReputationAlpha)
+		require.EqualValues(t, 0, d.Reputation.AuditReputationBeta)
+		require.EqualValues(t, 1, d.Reputation.UnknownAuditReputationAlpha)
+		require.EqualValues(t, 0, d.Reputation.UnknownAuditReputationBeta)
+	})
+}

satellite/overlay/service.go

@@ -155,6 +155,8 @@ const (
 	AuditFailure
 	// AuditUnknown represents an audit that resulted in an unknown error from the node.
 	AuditUnknown
+	// AuditOffline represents an audit where a node was offline.
+	AuditOffline
 )
 
 // UpdateRequest is used to update a node status.

satellite/overlay/suspension_test.go

@@ -247,10 +247,10 @@ func TestAuditSuspendDQDisabled(t *testing.T) {
 		require.Nil(t, n.UnknownAuditSuspended)
 		require.NotNil(t, n.Disqualified)
 
-		// offline node should still be suspended but not disqualified
+		// offline node should not be suspended or disqualified
 		n, err = oc.Get(ctx, offlineNodeID)
 		require.NoError(t, err)
-		require.NotNil(t, n.UnknownAuditSuspended)
+		require.Nil(t, n.UnknownAuditSuspended)
 		require.Nil(t, n.Disqualified)
 
 		// unknown node should still be suspended but not disqualified
@@ -342,7 +342,7 @@ func TestOfflineSuspend(t *testing.T) {
 
 		updateReq := &overlay.UpdateRequest{
 			NodeID:       nodeID,
-			AuditOutcome: overlay.AuditSuccess,
+			AuditOutcome: overlay.AuditOffline,
 			IsUp:         false,
 			AuditHistory: overlay.AuditHistoryConfig{
 				WindowSize:       time.Hour,

satellite/satellitedb/overlaycache.go

@@ -1246,8 +1246,11 @@ func (cache *overlaycache) populateUpdateNodeStats(dbNode *dbx.Node, updateReq *
 			updateReq.AuditWeight,
 			totalAuditCount,
 		)
-
+	case overlay.AuditOffline:
+		// for audit offline, only update total audit count
+		updatedTotalAuditCount = totalAuditCount + 1
 	}
+
 	mon.FloatVal("audit_reputation_alpha").Observe(auditAlpha)                //mon:locked
 	mon.FloatVal("audit_reputation_beta").Observe(auditBeta)                  //mon:locked
 	mon.FloatVal("unknown_audit_reputation_alpha").Observe(unknownAuditAlpha) //mon:locked