[2.5.1-auto] Recoverer: limit the number of pending payloads for an upkeep

core/services/ocr2/plugins/ocr2keeper/evm21/logprovider/recoverer.go

@@ -1,6 +1,7 @@
 package logprovider
 
 import (
+	"bytes"
 	"context"
 	"crypto/rand"
 	"errors"
@@ -43,6 +44,9 @@ var (
 	recoveryLogsBurst  = int64(500)
 	// blockTimeUpdateCadence is the cadence at which the chain's blocktime is re-calculated
 	blockTimeUpdateCadence = 10 * time.Minute
+	// maxPendingPayloadsPerUpkeep is the number of logs we can have pending for a single upkeep
+	// at any given time
+	maxPendingPayloadsPerUpkeep = 500
 )
 
 type LogRecoverer interface {
@@ -405,10 +409,14 @@ func (r *logRecoverer) recoverFilter(ctx context.Context, f upkeepFilter, startB
 	}
 	filteredLogs := r.filterFinalizedStates(f, logs, states)
 
-	added, alreadyPending := r.populatePending(f, filteredLogs)
+	added, alreadyPending, ok := r.populatePending(f, filteredLogs)
 	if added > 0 {
 		r.lggr.Debugw("found missed logs", "added", added, "alreadyPending", alreadyPending, "upkeepID", f.upkeepID)
 	}
+	if !ok {
+		r.lggr.Debugw("failed to add all logs to pending", "upkeepID", f.upkeepID)
+		return nil
+	}
 	r.filterStore.UpdateFilters(func(uf1, uf2 upkeepFilter) upkeepFilter {
 		uf1.lastRePollBlock = end
 		r.lggr.Debugw("Updated lastRePollBlock", "lastRePollBlock", end, "upkeepID", uf1.upkeepID)
@@ -419,13 +427,15 @@ func (r *logRecoverer) recoverFilter(ctx context.Context, f upkeepFilter, startB
 }
 
 // populatePending adds the logs to the pending list if they are not already pending.
-// returns the number of logs added and the number of logs that were already pending.
-func (r *logRecoverer) populatePending(f upkeepFilter, filteredLogs []logpoller.Log) (int, int) {
+// returns the number of logs added, the number of logs that were already pending,
+// and a flag that indicates whether some errors happened while we are trying to add to pending q.
+func (r *logRecoverer) populatePending(f upkeepFilter, filteredLogs []logpoller.Log) (int, int, bool) {
 	r.lock.Lock()
 	defer r.lock.Unlock()
 
 	pendingSizeBefore := len(r.pending)
 	alreadyPending := 0
+	errs := make([]error, 0)
 	for _, log := range filteredLogs {
 		trigger := logToTrigger(log)
 		// Set the checkBlock and Hash to zero so that the checkPipeline uses the latest block
@@ -453,13 +463,16 @@ func (r *logRecoverer) populatePending(f upkeepFilter, filteredLogs []logpoller.
 			continue
 		}
 		// r.lggr.Debugw("adding a payload to pending", "payload", payload)
-		r.visited[wid] = visitedRecord{
-			visitedAt: time.Now(),
-			payload:   payload,
+		if err := r.addPending(payload); err != nil {
+			errs = append(errs, err)
+		} else {
+			r.visited[wid] = visitedRecord{
+				visitedAt: time.Now(),
+				payload:   payload,
+			}
 		}
-		r.addPending(payload)
 	}
-	return len(r.pending) - pendingSizeBefore, alreadyPending
+	return len(r.pending) - pendingSizeBefore, alreadyPending, len(errs) == 0
 }
 
 // filterFinalizedStates filters out the log upkeeps that have already been completed (performed or ineligible).
@@ -619,9 +632,10 @@ func (r *logRecoverer) tryExpire(ctx context.Context, ids ...string) error {
 				removed++
 				continue
 			}
-			r.addPending(rec.payload)
-			rec.visitedAt = time.Now()
-			r.visited[ids[i]] = rec
+			if err := r.addPending(rec.payload); err == nil {
+				rec.visitedAt = time.Now()
+				r.visited[ids[i]] = rec
+			}
 		default:
 			delete(r.visited, ids[i])
 			removed++
@@ -637,17 +651,25 @@ func (r *logRecoverer) tryExpire(ctx context.Context, ids ...string) error {
 
 // addPending adds a payload to the pending list if it's not already there.
 // NOTE: the lock must be held before calling this function.
-func (r *logRecoverer) addPending(payload ocr2keepers.UpkeepPayload) {
+func (r *logRecoverer) addPending(payload ocr2keepers.UpkeepPayload) error {
 	var exist bool
 	pending := r.pending
+	upkeepPayloads := 0
 	for _, p := range pending {
+		if bytes.Equal(p.UpkeepID[:], payload.UpkeepID[:]) {
+			upkeepPayloads++
+		}
 		if p.WorkID == payload.WorkID {
 			exist = true
 		}
 	}
+	if upkeepPayloads >= maxPendingPayloadsPerUpkeep {
+		return fmt.Errorf("upkeep %v has too many payloads in pending queue", payload.UpkeepID)
+	}
 	if !exist {
 		r.pending = append(pending, payload)
 	}
+	return nil
 }
 
 // removePending removes a payload from the pending list.

core/services/ocr2/plugins/ocr2keeper/evm21/logprovider/recoverer_test.go

@@ -1093,40 +1093,75 @@ func TestLogRecoverer_GetProposalData(t *testing.T) {
 
 func TestLogRecoverer_pending(t *testing.T) {
 	tests := []struct {
-		name  string
-		exist []ocr2keepers.UpkeepPayload
-		new   []ocr2keepers.UpkeepPayload
-		want  []ocr2keepers.UpkeepPayload
+		name         string
+		maxPerUpkeep int
+		exist        []ocr2keepers.UpkeepPayload
+		new          []ocr2keepers.UpkeepPayload
+		errored      []bool
+		want         []ocr2keepers.UpkeepPayload
 	}{
 		{
-			"empty",
-			[]ocr2keepers.UpkeepPayload{},
-			[]ocr2keepers.UpkeepPayload{},
-			[]ocr2keepers.UpkeepPayload{},
+			name:         "empty",
+			maxPerUpkeep: 10,
+			exist:        []ocr2keepers.UpkeepPayload{},
+			new:          []ocr2keepers.UpkeepPayload{},
+			errored:      []bool{},
+			want:         []ocr2keepers.UpkeepPayload{},
 		},
 		{
-			"add new and existing",
-			[]ocr2keepers.UpkeepPayload{
+			name:         "add new and existing",
+			maxPerUpkeep: 10,
+			exist: []ocr2keepers.UpkeepPayload{
 				{WorkID: "1", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "1")},
 			},
-			[]ocr2keepers.UpkeepPayload{
+			new: []ocr2keepers.UpkeepPayload{
 				{WorkID: "1", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "1")},
 				{WorkID: "2", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "2")},
 			},
-			[]ocr2keepers.UpkeepPayload{
+			errored: []bool{false, false},
+			want: []ocr2keepers.UpkeepPayload{
 				{WorkID: "1", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "1")},
 				{WorkID: "2", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "2")},
 			},
 		},
+		{
+			name:         "exceed limits for upkeep",
+			maxPerUpkeep: 3,
+			exist: []ocr2keepers.UpkeepPayload{
+				{WorkID: "1", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "1")},
+				{WorkID: "2", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "1")},
+				{WorkID: "3", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "1")},
+			},
+			new: []ocr2keepers.UpkeepPayload{
+				{WorkID: "4", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "1")},
+			},
+			errored: []bool{true},
+			want: []ocr2keepers.UpkeepPayload{
+				{WorkID: "1", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "1")},
+				{WorkID: "2", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "1")},
+				{WorkID: "3", UpkeepID: core.GenUpkeepID(ocr2keepers.LogTrigger, "1")},
+			},
+		},
 	}
 
 	for _, tc := range tests {
 		t.Run(tc.name, func(t *testing.T) {
+			origMaxPendingPayloadsPerUpkeep := maxPendingPayloadsPerUpkeep
+			maxPendingPayloadsPerUpkeep = tc.maxPerUpkeep
+			defer func() {
+				maxPendingPayloadsPerUpkeep = origMaxPendingPayloadsPerUpkeep
+			}()
+
 			r := NewLogRecoverer(logger.TestLogger(t), nil, nil, nil, nil, nil, NewOptions(200))
 			r.lock.Lock()
 			r.pending = tc.exist
-			for _, p := range tc.new {
-				r.addPending(p)
+			for i, p := range tc.new {
+				err := r.addPending(p)
+				if tc.errored[i] {
+					require.Error(t, err)
+					continue
+				}
+				require.NoError(t, err)
 			}
 			pending := r.pending
 			require.GreaterOrEqual(t, len(pending), len(tc.new))