Merge #157933 #157953

157933: mma: fix minor todos related to pending changes r=wenyihu6 a=sumeerbhola

Some methods now take a PendingRangeChange as parameter, insted of a slice
of pendingReplicaChanges, to make it clearer that these are changes for
the same range.

Also removed two todos:
- The RangeID continues to be a field in pendingReplicaChange, since it is
  convenient (mainly for logging and testing), despite redundancy with the
  containing PendingRangeChange. This is because the contained relationship
  is not maintained in various maps keyed by the changeID.
- The changes in a rangeState are not modeled as a PendingRangeChange since
  the concept is redundant -- they are already contained in a range specific
  container.

Informs #157049

Epic: CRDB-55052

Release note: None

157953: release: released CockroachDB version v26.1.0-alpha.0. Next version: v26.1.0-alpha.1 r=celiala a=cockroach-teamcity


Release note: None
Epic: None
Release justification: non-production (release infra) change.


Co-authored-by: sumeerbhola <[email protected]>
Co-authored-by: Justin Beaver <[email protected]>

Author: 3 people

pkg/build/version.txt

@@ -1 +1 @@
-v26.1.0-alpha.00000000
+v26.1.0-alpha.1

pkg/kv/kvserver/allocator/mmaprototype/allocator_state.go

@@ -291,7 +291,10 @@ func (a *allocatorState) AdjustPendingChangeDisposition(change PendingRangeChang
 	}
 	if !success {
 		// Check that we can undo these changes.
-		if err := a.cs.preCheckOnUndoReplicaChanges(changes); err != nil {
+		if err := a.cs.preCheckOnUndoReplicaChanges(PendingRangeChange{
+			RangeID:               change.RangeID,
+			pendingReplicaChanges: changes,
+		}); err != nil {
 			panic(err)
 		}
 	}
@@ -308,7 +311,7 @@ func (a *allocatorState) AdjustPendingChangeDisposition(change PendingRangeChang
 func (a *allocatorState) RegisterExternalChange(change PendingRangeChange) (ok bool) {
 	a.mu.Lock()
 	defer a.mu.Unlock()
-	if err := a.cs.preCheckOnApplyReplicaChanges(change.pendingReplicaChanges); err != nil {
+	if err := a.cs.preCheckOnApplyReplicaChanges(change); err != nil {
 		a.mmaMetrics.ExternalFailedToRegister.Inc(1)
 		log.KvDistribution.Infof(context.Background(),
 			"did not register external changes: due to %v", err)

pkg/kv/kvserver/allocator/mmaprototype/cluster_state.go

@@ -169,7 +169,10 @@ type ReplicaChange struct {
 
 	// target is the target {store,node} for the change.
 	target roachpb.ReplicationTarget
-	// TODO(sumeer): remove rangeID.
+	// rangeID is the same as that in the PendingRangeChange.RangeID this change
+	// is part of. It is duplicated here since the individual
+	// pendingReplicaChanges are kept in various maps keyed by the changeID, and
+	// having the RangeID field on each change is convenient.
 	rangeID roachpb.RangeID
 
 	// NB: 0 is not a valid ReplicaID, but this component does not care about
@@ -1200,8 +1203,6 @@ type rangeState struct {
 	// Life-cycle matches clusterState.pendingChanges. The consolidated
 	// rangeState.pendingChanges across all ranges in clusterState.ranges will
 	// be identical to clusterState.pendingChanges.
-	//
-	// TODO(sumeer): replace by PendingRangeChange.
 	pendingChanges []*pendingReplicaChange
 	// When set, the pendingChanges can not be rolled back anymore. They have
 	// to be enacted, or discarded wholesale in favor of the latest RangeMsg
@@ -1607,7 +1608,10 @@ func (cs *clusterState) processStoreLeaseholderMsgInternal(
 				// remainingChanges, but they are not the same slice.
 				rc := rs.pendingChanges
 				rs.pendingChanges = nil
-				err := cs.preCheckOnApplyReplicaChanges(remainingChanges)
+				err := cs.preCheckOnApplyReplicaChanges(PendingRangeChange{
+					RangeID:               rangeMsg.RangeID,
+					pendingReplicaChanges: remainingChanges,
+				})
 				valid = err == nil
 				if err != nil {
 					reason = redact.Sprint(err)
@@ -1907,7 +1911,10 @@ func (cs *clusterState) gcPendingChanges(now time.Time) {
 		if !startTime.Add(pendingChangeGCDuration).Before(now) {
 			continue
 		}
-		if err := cs.preCheckOnUndoReplicaChanges(rs.pendingChanges); err != nil {
+		if err := cs.preCheckOnUndoReplicaChanges(PendingRangeChange{
+			RangeID:               rangeID,
+			pendingReplicaChanges: rs.pendingChanges,
+		}); err != nil {
 			panic(err)
 		}
 		// Gather the changeIDs, since calls to undoPendingChange modify the
@@ -1980,24 +1987,6 @@ func printMapPendingChanges(changes map[changeID]*pendingReplicaChange) string {
 	return buf.String()
 }
 
-//lint:ignore U1000 used in tests
-func printPendingChanges(changes []*pendingReplicaChange) string {
-	var buf strings.Builder
-	fmt.Fprintf(&buf, "pending(%d)", len(changes))
-	for _, change := range changes {
-		fmt.Fprintf(&buf, "\nchange-id=%d store-id=%v node-id=%v range-id=%v load-delta=%v start=%v",
-			change.changeID, change.target.StoreID, change.target.NodeID, change.rangeID,
-			change.loadDelta, change.startTime,
-		)
-		if !(change.enactedAtTime == time.Time{}) {
-			fmt.Fprintf(&buf, " enacted=%v",
-				change.enactedAtTime)
-		}
-		fmt.Fprintf(&buf, "\n  prev=(%v)\n  next=(%v)", change.prev, change.next)
-	}
-	return buf.String()
-}
-
 // addPendingRangeChange takes a range change containing a set of replica
 // changes, and applies the changes as pending. The application updates the
 // adjusted load, tracked pending changes and changeIDs to reflect the pending
@@ -2062,16 +2051,11 @@ func (cs *clusterState) addPendingRangeChange(change PendingRangeChange) {
 // been added at a store, but it has not yet received the lease. Finally, it
 // checks that after the changes are applied there is exactly one leaseholder.
 // It returns a non-nil error if any of these checks fail.
-//
-// REQUIRES: all the changes are to the same range, and there is at most one
-// change per store.
-//
-// TODO(sumeer): change to take PendingRangeChange as parameter
-func (cs *clusterState) preCheckOnApplyReplicaChanges(changes []*pendingReplicaChange) error {
-	if len(changes) == 0 {
+func (cs *clusterState) preCheckOnApplyReplicaChanges(rangeChange PendingRangeChange) error {
+	if len(rangeChange.pendingReplicaChanges) == 0 {
 		return nil
 	}
-	rangeID := changes[0].rangeID
+	rangeID := rangeChange.RangeID
 	curr, ok := cs.ranges[rangeID]
 	// Return early if range already has some pending changes or the range does not exist.
 	if !ok {
@@ -2086,7 +2070,7 @@ func (cs *clusterState) preCheckOnApplyReplicaChanges(changes []*pendingReplicaC
 	copiedCurr := rangeState{
 		replicas: append([]StoreIDAndReplicaState{}, curr.replicas...),
 	}
-	for _, change := range changes {
+	for _, change := range rangeChange.pendingReplicaChanges {
 		// Check that all changes correspond to the same range. Panic otherwise.
 		if change.rangeID != rangeID {
 			panic(errors.AssertionFailedf("unexpected change rangeID %d != %d", change.rangeID, rangeID))
@@ -2115,20 +2099,17 @@ func (cs *clusterState) preCheckOnApplyReplicaChanges(changes []*pendingReplicaC
 // preCheckOnUndoReplicaChanges does some validation of the changes being
 // proposed for undo.
 //
-// REQUIRES: changes is non-empty; all changes are to the same range; the
-// rangeState.pendingChangeNoRollback is false.
+// REQUIRES: the rangeState.pendingChangeNoRollback is false.
 //
 // This method is defensive since if we always check against the current state
 // before allowing a change to be added (including re-addition after a
 // StoreLeaseholderMsg), we should never have invalidity during an undo, if
 // all the changes are being undone.
-//
-// TODO(sumeer): change to take PendingRangeChange as parameter
-func (cs *clusterState) preCheckOnUndoReplicaChanges(changes []*pendingReplicaChange) error {
-	if len(changes) == 0 {
-		panic(errors.AssertionFailedf("no changes to undo"))
+func (cs *clusterState) preCheckOnUndoReplicaChanges(rangeChange PendingRangeChange) error {
+	if len(rangeChange.pendingReplicaChanges) == 0 {
+		return nil
 	}
-	rangeID := changes[0].rangeID
+	rangeID := rangeChange.RangeID
 	curr, ok := cs.ranges[rangeID]
 	if !ok {
 		return errors.Errorf("range %v does not exist in cluster state", rangeID)
@@ -2137,7 +2118,7 @@ func (cs *clusterState) preCheckOnUndoReplicaChanges(changes []*pendingReplicaCh
 	copiedCurr := &rangeState{
 		replicas: append([]StoreIDAndReplicaState{}, curr.replicas...),
 	}
-	for _, change := range changes {
+	for _, change := range rangeChange.pendingReplicaChanges {
 		if change.rangeID != rangeID {
 			panic(errors.AssertionFailedf("unexpected change rangeID %d != %d", change.rangeID, rangeID))
 		}

pkg/kv/kvserver/allocator/mmaprototype/cluster_state_rebalance_stores.go

@@ -413,7 +413,7 @@ func (re *rebalanceEnv) rebalanceReplicas(
 		replicaChanges := makeRebalanceReplicaChanges(
 			rangeID, rstate.replicas, rstate.load, addTarget, removeTarget)
 		rangeChange := MakePendingRangeChange(rangeID, replicaChanges[:])
-		if err = re.preCheckOnApplyReplicaChanges(rangeChange.pendingReplicaChanges); err != nil {
+		if err = re.preCheckOnApplyReplicaChanges(rangeChange); err != nil {
 			panic(errors.Wrapf(err, "pre-check failed for replica changes: %v for %v",
 				replicaChanges, rangeID))
 		}
@@ -591,7 +591,7 @@ func (re *rebalanceEnv) rebalanceLeases(
 		replicaChanges := MakeLeaseTransferChanges(
 			rangeID, rstate.replicas, rstate.load, addTarget, removeTarget)
 		leaseChange := MakePendingRangeChange(rangeID, replicaChanges[:])
-		if err := re.preCheckOnApplyReplicaChanges(leaseChange.pendingReplicaChanges); err != nil {
+		if err := re.preCheckOnApplyReplicaChanges(leaseChange); err != nil {
 			panic(errors.Wrapf(err, "pre-check failed for lease transfer %v", leaseChange))
 		}
 		re.addPendingRangeChange(leaseChange)