*: consensus abstraction

app/app.go

@@ -47,7 +47,8 @@
 	"github.com/obolnetwork/charon/core/aggsigdb"
 	"github.com/obolnetwork/charon/core/bcast"
 	"github.com/obolnetwork/charon/core/consensus"
-	pbv1 "github.com/obolnetwork/charon/core/corepb/v1"
+	"github.com/obolnetwork/charon/core/consensus/protocols"
+	"github.com/obolnetwork/charon/core/consensus/qbft"
 	"github.com/obolnetwork/charon/core/dutydb"
 	"github.com/obolnetwork/charon/core/fetcher"
 	"github.com/obolnetwork/charon/core/infosync"
@@ -92,6 +93,7 @@
 	SimnetBMockFuzz         bool
 	TestnetConfig           eth2util.Network
 	ProcDirectory           string
+	ConsensusProtocol       string
 
 	TestConfig TestConfig
 }
@@ -257,8 +259,6 @@
 
 	wirePeerInfo(life, tcpNode, peerIDs, cluster.GetInitialMutationHash(), sender, conf.BuilderAPI)
 
-	consensusDebugger := consensus.NewDebugger()
-
 	// seenPubkeys channel to send seen public keys from validatorapi to monitoringapi.
 	seenPubkeys := make(chan core.PubKey)
 	seenPubkeysFunc := func(pk core.PubKey) {
@@ -281,6 +281,8 @@
 		return err
 	}
 
+	consensusDebugger := consensus.NewDebugger()
+
 	wireMonitoringAPI(ctx, life, conf.MonitoringAddr, conf.DebugAddr, tcpNode, eth2Cl, peerIDs,
 		promRegistry, consensusDebugger, pubkeys, seenPubkeys, vapiCalls, len(cluster.GetValidators()))
 
@@ -526,14 +528,20 @@
 
 	retryer := retry.New[core.Duty](deadlineFunc)
 
-	cons, startCons, err := newConsensus(cluster, tcpNode, p2pKey, sender,
-		deadlinerFunc("consensus"), gaterFunc, consensusDebugger.AddInstance)
+	consensusFactory, err := consensus.NewConsensusFactory(tcpNode, sender, peers, p2pKey, deadlinerFunc, gaterFunc, consensusDebugger)
 	if err != nil {
 		return err
 	}
 
+	defaultConsensus := consensusFactory.DefaultConsensus()
+	startDefaultConsensus := lifecycle.HookFuncCtx(defaultConsensus.Start)
+
+	coreConsensus := consensusFactory.CurrentConsensus() // points to DefaultConsensus() initially
+
+	// Priority protocol always uses QBFTv2.
 	err = wirePrioritise(ctx, conf, life, tcpNode, peerIDs, int(cluster.GetThreshold()),
-		sender.SendReceive, cons, sched, p2pKey, deadlineFunc)
+		sender.SendReceive, defaultConsensus, sched, p2pKey, deadlineFunc,
+		consensusFactory, cluster.GetConsensusProtocol())
 	if err != nil {
 		return err
 	}
@@ -553,12 +561,13 @@
 		return err
 	}
 
+	// Core always uses the "current" consensus that is changed dynamically.
 	opts := []core.WireOption{
 		core.WithTracing(),
 		core.WithTracking(track, inclusion),
 		core.WithAsyncRetry(retryer),
 	}
-	core.Wire(sched, fetch, cons, dutyDB, vapi, parSigDB, parSigEx, sigAgg, aggSigDB, broadcaster, opts...)
+	core.Wire(sched, fetch, coreConsensus, dutyDB, vapi, parSigDB, parSigEx, sigAgg, aggSigDB, broadcaster, opts...)
 
 	err = wireValidatorMock(ctx, conf, eth2Cl, pubshares, sched)
 	if err != nil {
@@ -570,7 +579,7 @@
 	}
 
 	life.RegisterStart(lifecycle.AsyncBackground, lifecycle.StartScheduler, lifecycle.HookFuncErr(sched.Run))
-	life.RegisterStart(lifecycle.AsyncAppCtx, lifecycle.StartP2PConsensus, startCons)
+	life.RegisterStart(lifecycle.AsyncAppCtx, lifecycle.StartP2PConsensus, startDefaultConsensus)
 	life.RegisterStart(lifecycle.AsyncAppCtx, lifecycle.StartAggSigDB, lifecycle.HookFuncCtx(aggSigDB.Run))
 	life.RegisterStart(lifecycle.AsyncAppCtx, lifecycle.StartParSigDB, lifecycle.HookFuncCtx(parSigDB.Trim))
 	life.RegisterStart(lifecycle.AsyncAppCtx, lifecycle.StartTracker, lifecycle.HookFuncCtx(inclusion.Run))
@@ -585,8 +594,9 @@
 func wirePrioritise(ctx context.Context, conf Config, life *lifecycle.Manager, tcpNode host.Host,
 	peers []peer.ID, threshold int, sendFunc p2p.SendReceiveFunc, coreCons core.Consensus,
 	sched core.Scheduler, p2pKey *k1.PrivateKey, deadlineFunc func(duty core.Duty) (time.Time, bool),
+	consensusFactory core.ConsensusFactory, clusterPreferredProtocol string,
 ) error {
-	cons, ok := coreCons.(*consensus.Component)
+	cons, ok := coreCons.(*qbft.Consensus)
 	if !ok {
 		// Priority protocol not supported for leader cast.
 		return nil
@@ -602,9 +612,22 @@
 		return err
 	}
 
+	// The initial protocols order as defined by implementation is altered by:
+	// 1. Prioritizing the cluster (lock) preferred protocol to the top.
+	// 2. Prioritizing the protocol specified by CLI flag (cluster run) to the top.
+	// In all cases this prioritizes all versions of the protocol identified by name.
+	// The order of all these operations are important.
+	allProtocols := Protocols()
+	if clusterPreferredProtocol != "" {
+		allProtocols = protocols.PrioritizeProtocolsByName(clusterPreferredProtocol, allProtocols)
+	}
+	if conf.ConsensusProtocol != "" {
+		allProtocols = protocols.PrioritizeProtocolsByName(conf.ConsensusProtocol, allProtocols)
+	}
+
 	isync := infosync.New(prio,
 		version.Supported(),
-		Protocols(),
+		allProtocols,
 		ProposalTypes(conf.BuilderAPI, conf.SyntheticBlockProposals),
 	)
 
@@ -621,6 +644,25 @@
 		prio.Subscribe(conf.TestConfig.PrioritiseCallback)
 	}
 
+	prio.Subscribe(func(ctx context.Context, _ core.Duty, tr []priority.TopicResult) error {
+		for _, t := range tr {
+			if t.Topic == infosync.TopicProtocol {
+				allProtocols := t.PrioritiesOnly()
+				preferredConsensusProtocol := protocols.MostPreferredConsensusProtocol(allProtocols)
+
+				if err := consensusFactory.SetCurrentConsensusForProtocol(protocol.ID(preferredConsensusProtocol)); err != nil {
+					log.Error(ctx, "Failed to set current consensus protocol", err, z.Str("protocol", preferredConsensusProtocol))
+				} else {
+					log.Info(ctx, "Current consensus protocol changed", z.Str("protocol", preferredConsensusProtocol))
+				}
+
+				break
+			}
+		}
+
+		return nil
+	})
+
 	life.RegisterStart(lifecycle.AsyncAppCtx, lifecycle.StartPeerInfo, lifecycle.HookFuncCtx(prio.Start))
 
 	return nil
@@ -918,24 +960,6 @@
 	return eth2Cl, nil
 }
 
-// newConsensus returns a new consensus component and its start lifecycle hook.
-func newConsensus(cluster *manifestpb.Cluster, tcpNode host.Host, p2pKey *k1.PrivateKey,
-	sender *p2p.Sender, deadliner core.Deadliner, gaterFunc core.DutyGaterFunc,
-	qbftSniffer func(*pbv1.SniffedConsensusInstance),
-) (core.Consensus, lifecycle.IHookFunc, error) {
-	peers, err := manifest.ClusterPeers(cluster)
-	if err != nil {
-		return nil, nil, err
-	}
-
-	comp, err := consensus.New(tcpNode, sender, peers, p2pKey, deadliner, gaterFunc, qbftSniffer)
-	if err != nil {
-		return nil, nil, err
-	}
-
-	return comp, lifecycle.HookFuncCtx(comp.Start), nil
-}
-
 // createMockValidators creates mock validators identified by their public shares.
 func createMockValidators(pubkeys []eth2p0.BLSPubKey) beaconmock.ValidatorSet {
 	resp := make(beaconmock.ValidatorSet)
@@ -1079,7 +1103,7 @@
 // Protocols returns the list of supported Protocols in order of precedence.
 func Protocols() []protocol.ID {
 	var resp []protocol.ID
-	resp = append(resp, consensus.Protocols()...)
+	resp = append(resp, protocols.Protocols()...)
 	resp = append(resp, parsigex.Protocols()...)
 	resp = append(resp, peerinfo.Protocols()...)
 	resp = append(resp, priority.Protocols()...)

cluster/manifest/mutationlegacylock.go

@@ -145,12 +145,13 @@ func transformLegacyLock(input *manifestpb.Cluster, signed *manifestpb.SignedMut
 	}
 
 	return &manifestpb.Cluster{
-		Name:         lock.Name,
-		Threshold:    int32(lock.Threshold),
-		DkgAlgorithm: lock.DKGAlgorithm,
-		ForkVersion:  lock.ForkVersion,
-		Validators:   vals,
-		Operators:    ops,
+		Name:              lock.Name,
+		Threshold:         int32(lock.Threshold),
+		DkgAlgorithm:      lock.DKGAlgorithm,
+		ForkVersion:       lock.ForkVersion,
+		ConsensusProtocol: lock.ConsensusProtocol,
+		Validators:        vals,
+		Operators:         ops,
 	}, nil
 }