fix: Publisher-mode synchronization option for failover scenario

CHANGELOG.md

@@ -13,7 +13,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Changes
 
-- Improve P2P transient network failure [#3212](https://github.com/evstack/ev-node/pull/3212)
+* Added publisher-mode synchronization option for failover scenarios with early P2P infrastructure readiness [#3222](https://github.com/evstack/ev-node/pull/3222)
+* Improve P2P transient network failure [#3212](https://github.com/evstack/ev-node/pull/3212)
 * Improve execution/evm check for stored meta not stale [#3221](https://github.com/evstack/ev-node/pull/3221)
 
 ## v1.1.0-rc.1

apps/testapp/go.mod

@@ -79,7 +79,7 @@ require (
 	github.com/googleapis/gax-go/v2 v2.20.0 // indirect
 	github.com/gorilla/websocket v1.5.3 // indirect
 	github.com/grpc-ecosystem/grpc-gateway/v2 v2.28.0 // indirect
-	github.com/hashicorp/go-hclog v1.6.2 // indirect
+	github.com/hashicorp/go-hclog v1.6.3 // indirect
 	github.com/hashicorp/go-immutable-radix v1.3.1 // indirect
 	github.com/hashicorp/go-metrics v0.5.4 // indirect
 	github.com/hashicorp/go-msgpack v0.5.5 // indirect

apps/testapp/go.sum

@@ -650,8 +650,9 @@ github.com/hashicorp/consul/sdk v0.1.1/go.mod h1:VKf9jXwCTEY1QZP2MOLRhb5i/I/ssyN
 github.com/hashicorp/errwrap v1.0.0/go.mod h1:YH+1FKiLXxHSkmPseP+kNlulaMuP3n2brvKWEqk/Jc4=
 github.com/hashicorp/go-cleanhttp v0.5.0/go.mod h1:JpRdi6/HCYpAwUzNwuwqhbovhLtngrth3wmdIIUrZ80=
 github.com/hashicorp/go-cleanhttp v0.5.1/go.mod h1:JpRdi6/HCYpAwUzNwuwqhbovhLtngrth3wmdIIUrZ80=
-github.com/hashicorp/go-hclog v1.6.2 h1:NOtoftovWkDheyUM/8JW3QMiXyxJK3uHRK7wV04nD2I=
 github.com/hashicorp/go-hclog v1.6.2/go.mod h1:W4Qnvbt70Wk/zYJryRzDRU/4r0kIg0PVHBcfoyhpF5M=
+github.com/hashicorp/go-hclog v1.6.3 h1:Qr2kF+eVWjTiYmU7Y31tYlP1h0q/X3Nl3tPGdaB11/k=
+github.com/hashicorp/go-hclog v1.6.3/go.mod h1:W4Qnvbt70Wk/zYJryRzDRU/4r0kIg0PVHBcfoyhpF5M=
 github.com/hashicorp/go-immutable-radix v1.0.0/go.mod h1:0y9vanUI8NX6FsYoO3zeMjhV/C5i9g4Q3DwcSNZ4P60=
 github.com/hashicorp/go-immutable-radix v1.3.1 h1:DKHmCUm2hRBK510BaiZlwvpD40f8bJFeZnpfm2KLowc=
 github.com/hashicorp/go-immutable-radix v1.3.1/go.mod h1:0y9vanUI8NX6FsYoO3zeMjhV/C5i9g4Q3DwcSNZ4P60=

block/internal/syncing/syncer.go

@@ -694,6 +694,12 @@ var (
 // TrySyncNextBlock attempts to sync the next available block
 // the event is always the next block in sequence as processHeightEvent ensures it.
 func (s *Syncer) TrySyncNextBlock(ctx context.Context, event *common.DAHeightEvent) error {
+	return s.trySyncNextBlockWithState(ctx, event, s.getLastState())
+}
+
+// trySyncNextBlockWithState attempts to sync the next available block using
+// the provided current state as the validation/apply baseline.
+func (s *Syncer) trySyncNextBlockWithState(ctx context.Context, event *common.DAHeightEvent, currentState types.State) error {
 	select {
 	case <-ctx.Done():
 		return ctx.Err()
@@ -703,7 +709,6 @@ func (s *Syncer) TrySyncNextBlock(ctx context.Context, event *common.DAHeightEve
 	header := event.Header
 	data := event.Data
 	nextHeight := event.Header.Height()
-	currentState := s.getLastState()
 	headerHash := header.Hash().String()
 
 	s.logger.Info().Uint64("height", nextHeight).Str("source", string(event.Source)).Msg("syncing block")
@@ -1201,6 +1206,7 @@ func (s *Syncer) RecoverFromRaft(ctx context.Context, raftState *raft.RaftBlockS
 			s.logger.Debug().Err(err).Msg("no state in store, using genesis defaults for recovery")
 			currentState = types.State{
 				ChainID:         s.genesis.ChainID,
+				InitialHeight:   s.genesis.InitialHeight,
 				LastBlockHeight: s.genesis.InitialHeight - 1,
 			}
 		}
@@ -1214,11 +1220,12 @@ func (s *Syncer) RecoverFromRaft(ctx context.Context, raftState *raft.RaftBlockS
 		return nil
 	} else if currentState.LastBlockHeight+1 == raftState.Height { // raft is 1 block ahead
 		// apply block
-		err := s.TrySyncNextBlock(ctx, &common.DAHeightEvent{
+		event := &common.DAHeightEvent{
 			Header: &header,
 			Data:   &data,
 			Source: "",
-		})
+		}
+		err := s.trySyncNextBlockWithState(ctx, event, currentState)
 		if err != nil {
 			return err
 		}

block/internal/syncing/syncer_test.go

@@ -26,6 +26,7 @@ import (
 	"github.com/evstack/ev-node/pkg/config"
 	datypes "github.com/evstack/ev-node/pkg/da/types"
 	"github.com/evstack/ev-node/pkg/genesis"
+	"github.com/evstack/ev-node/pkg/raft"
 	signerpkg "github.com/evstack/ev-node/pkg/signer"
 	"github.com/evstack/ev-node/pkg/signer/noop"
 	"github.com/evstack/ev-node/pkg/store"
@@ -306,6 +307,121 @@ func TestSequentialBlockSync(t *testing.T) {
 	requireEmptyChan(t, errChan)
 }
 
+func TestSyncer_RecoverFromRaft_BootstrapsStateWhenUninitialized(t *testing.T) {
+	ds := dssync.MutexWrap(datastore.NewMapDatastore())
+	st := store.New(ds)
+
+	cm, err := cache.NewManager(config.DefaultConfig(), st, zerolog.Nop())
+	require.NoError(t, err)
+
+	addr, pub, signer := buildSyncTestSigner(t)
+	gen := genesis.Genesis{
+		ChainID:         "1234",
+		InitialHeight:   1,
+		StartTime:       time.Now().Add(-time.Second),
+		ProposerAddress: addr,
+	}
+
+	mockExec := testmocks.NewMockExecutor(t)
+	mockHeaderStore := extmocks.NewMockStore[*types.P2PSignedHeader](t)
+	mockDataStore := extmocks.NewMockStore[*types.P2PData](t)
+	s := NewSyncer(
+		st,
+		mockExec,
+		nil,
+		cm,
+		common.NopMetrics(),
+		config.DefaultConfig(),
+		gen,
+		mockHeaderStore,
+		mockDataStore,
+		zerolog.Nop(),
+		common.DefaultBlockOptions(),
+		make(chan error, 1),
+		nil,
+	)
+
+	// lastState intentionally not initialized to simulate recovery-before-start path.
+	data := makeData(gen.ChainID, 1, 0)
+	headerBz, hdr := makeSignedHeaderBytes(t, gen.ChainID, 1, addr, pub, signer, []byte("app0"), data, nil)
+	dataBz, err := data.MarshalBinary()
+	require.NoError(t, err)
+
+	mockExec.EXPECT().
+		ExecuteTxs(mock.Anything, mock.Anything, uint64(1), mock.Anything, mock.Anything).
+		Return([]byte("app1"), nil).
+		Once()
+
+	err = s.RecoverFromRaft(t.Context(), &raft.RaftBlockState{
+		Height: 1,
+		Hash:   hdr.Hash(),
+		Header: headerBz,
+		Data:   dataBz,
+	})
+	require.NoError(t, err)
+
+	state, err := st.GetState(t.Context())
+	require.NoError(t, err)
+	require.Equal(t, gen.ChainID, state.ChainID)
+	require.Equal(t, uint64(1), state.LastBlockHeight)
+}
+
+func TestSyncer_RecoverFromRaft_KeepsStrictValidationAfterStateExists(t *testing.T) {
+	ds := dssync.MutexWrap(datastore.NewMapDatastore())
+	st := store.New(ds)
+
+	cm, err := cache.NewManager(config.DefaultConfig(), st, zerolog.Nop())
+	require.NoError(t, err)
+
+	addr, pub, signer := buildSyncTestSigner(t)
+	gen := genesis.Genesis{
+		ChainID:         "1234",
+		InitialHeight:   1,
+		StartTime:       time.Now().Add(-time.Second),
+		ProposerAddress: addr,
+	}
+
+	mockExec := testmocks.NewMockExecutor(t)
+	mockHeaderStore := extmocks.NewMockStore[*types.P2PSignedHeader](t)
+	mockDataStore := extmocks.NewMockStore[*types.P2PData](t)
+	s := NewSyncer(
+		st,
+		mockExec,
+		nil,
+		cm,
+		common.NopMetrics(),
+		config.DefaultConfig(),
+		gen,
+		mockHeaderStore,
+		mockDataStore,
+		zerolog.Nop(),
+		common.DefaultBlockOptions(),
+		make(chan error, 1),
+		nil,
+	)
+
+	// Non-empty state must remain strictly validated.
+	s.SetLastState(types.State{
+		ChainID:         "wrong-chain",
+		InitialHeight:   1,
+		LastBlockHeight: 0,
+	})
+
+	data := makeData(gen.ChainID, 1, 0)
+	headerBz, hdr := makeSignedHeaderBytes(t, gen.ChainID, 1, addr, pub, signer, []byte("app0"), data, nil)
+	dataBz, err := data.MarshalBinary()
+	require.NoError(t, err)
+
+	err = s.RecoverFromRaft(t.Context(), &raft.RaftBlockState{
+		Height: 1,
+		Hash:   hdr.Hash(),
+		Header: headerBz,
+		Data:   dataBz,
+	})
+	require.Error(t, err)
+	require.ErrorContains(t, err, "invalid chain ID")
+}
+
 func TestSyncer_processPendingEvents(t *testing.T) {
 	ds := dssync.MutexWrap(datastore.NewMapDatastore())
 	st := store.New(ds)

docs/guides/raft_production.md

@@ -22,7 +22,7 @@ This guide details the Raft consensus implementation in `ev-node`, used for High
 
 ## Configuration
 
-Raft is configured via CLI flags or the `config.toml` file under the `[raft]` (or `[rollkit.raft]`) section.
+Raft is configured via CLI flags or the `config.toml` file under the `[raft]` (or `[evnode.raft]`) section.
 
 ### Essential Flags
 
@@ -33,7 +33,7 @@ Raft is configured via CLI flags or the `config.toml` file under the `[raft]` (o
 | `--evnode.raft.raft_addr` | `raft.raft_addr` | TCP address for Raft transport. | `0.0.0.0:5001` (Bind to private IP) |
 | `--evnode.raft.raft_dir` | `raft.raft_dir` | Directory for Raft data. | `/data/raft` (Must be persistent) |
 | `--evnode.raft.peers` | `raft.peers` | Comma-separated list of peer addresses in format `nodeID@host:port`. | `[email protected]:5001,[email protected]:5001,[email protected]:5001` |
-| `--evnode.raft.bootstrap` | `raft.bootstrap` | Bootstrap the cluster. **Required** for initial setup. | `true` (See Limitations) |
+| `--evnode.raft.bootstrap` | `raft.bootstrap` | Compatibility flag. Startup mode is selected automatically from persisted raft configuration state. | optional |
 
 ### Timeout Tuning
 
@@ -55,11 +55,15 @@ Ideally, a failover should complete within `2 * BlockTime` to minimize user impa
 
 ## Production Deployment Principles
 
-### 1. Static Peering & Bootstrap
-Current implementation requires **Bootstrap Mode** (`--evnode.raft.bootstrap=true`) for all nodes participating in the cluster initialization.
-*   **All nodes** should list the full set of peers in `--evnode.raft.peers`.
+### 1. Static Peering & Automatic Startup Mode
+Use static peering with automatic mode selection from local raft configuration:
+*   If local raft configuration already exists in `--evnode.raft.raft_dir`, the node starts in rejoin mode.
+*   If no local raft configuration exists yet, the node bootstraps from configured peers.
+*   `--evnode.raft.bootstrap` is retained for compatibility but does not control mode selection.
+*   **All configured cluster members** should list the full set of peers in `--evnode.raft.peers`.
 *   The `peers` list format is strict: `NodeID@Host:Port`.
-*   **Limitation**: Dynamic addition of peers (Run-time Membership Changes) via RPC/CLI is not currently exposed. The cluster membership is static based on the initial bootstrap configuration.
+*   **Limitation**: Dynamic addition of peers (run-time membership changes) via RPC/CLI is not currently exposed.
+*   **Not supported**: Joining an existing cluster as a brand-new node that was not part of the initial static membership.
 
 ### 2. Infrastructure Requirements
 *   **Encrypted Network (CRITICAL)**: Raft traffic is **unencrypted** (plain TCP). You **MUST** run the cluster inside a private network, VPN, or encrypted mesh (e.g., WireGuard, Tailscale). **Never expose Raft ports to the public internet**; doing so allows attackers to hijack the cluster consensus.
@@ -86,13 +90,13 @@ Monitor the following metrics (propagated via Prometheus if enabled):
 
 ```bash
 ./ev-node start \
-  --node.aggregator \
-  --raft.enable \
-  --raft.node_id="node-1" \
-  --raft.raft_addr="0.0.0.0:5001" \
-  --raft.raft_dir="/var/lib/ev-node/raft" \
-  --raft.bootstrap=true \
-  --raft.peers="[email protected]:5001,[email protected]:5001,[email protected]:5001" \
-  --p2p.listen_address="/ip4/0.0.0.0/tcp/26656" \
+  --evnode.node.aggregator=true \
+  --evnode.raft.enable=true \
+  --evnode.raft.node_id="node-1" \
+  --evnode.raft.raft_addr="0.0.0.0:5001" \
+  --evnode.raft.raft_dir="/var/lib/ev-node/raft" \
+  --evnode.raft.bootstrap=true \
+  --evnode.raft.peers="[email protected]:5001,[email protected]:5001,[email protected]:5001" \
+  --evnode.p2p.listen_address="/ip4/0.0.0.0/tcp/26656" \
   ...other flags
 ```

docs/learn/config.md

@@ -1321,7 +1321,7 @@ _Constant:_ `FlagRaftDir`
 ### Raft Bootstrap
 
 **Description:**
-If true, bootstraps a new Raft cluster. Only set this on the very first node when initializing a new cluster.
+Legacy compatibility flag. Startup mode is now auto-selected from persisted raft configuration state, so this flag is not used to choose bootstrap vs rejoin.
 
 **YAML:**
 
@@ -1352,6 +1352,16 @@ raft:
 _Default:_ `""` (empty)
 _Constant:_ `FlagRaftPeers`
 
+### Raft Startup Mode
+
+Raft startup mode is selected automatically from local raft configuration state:
+
+* If the node already has persisted raft configuration in `raft.raft_dir`, it starts in rejoin mode.
+* If no raft configuration exists yet, it bootstraps a cluster from configured peers.
+* `raft.bootstrap` is retained for compatibility but does not control mode selection.
+
+`--evnode.raft.rejoin` has been removed.
+
 ### Raft Snap Count
 
 **Description:**

node/failover.go

@@ -33,6 +33,8 @@ type failoverState struct {
 	dataSyncService   *evsync.DataSyncService
 	rpcServer         *http.Server
 	bc                *block.Components
+	raftNode          *raft.Node
+	isAggregator      bool
 
 	// catchup fields — used when the aggregator needs to sync before producing
 	catchupEnabled bool
@@ -172,13 +174,41 @@ func setupFailoverState(
 		dataSyncService:   dataSyncService,
 		rpcServer:         rpcServer,
 		bc:                bc,
+		raftNode:          raftNode,
+		isAggregator:      isAggregator,
 		store:             rktStore,
 		catchupEnabled:    catchupEnabled,
 		catchupTimeout:    nodeConfig.Node.CatchupTimeout.Duration,
 		daBlockTime:       nodeConfig.DA.BlockTime.Duration,
 	}, nil
 }
 
+// shouldStartSyncInPublisherMode avoids startup deadlock when a raft leader boots
+// with empty sync stores and no peer can serve height 1 yet.
+func (f *failoverState) shouldStartSyncInPublisherMode(ctx context.Context) bool {
+	if !f.isAggregator || f.raftNode == nil || !f.raftNode.IsLeader() {
+		return false
+	}
+
+	storeHeight, err := f.store.Height(ctx)
+	if err != nil {
+		f.logger.Warn().Err(err).Msg("cannot determine store height; keeping blocking sync startup")
+		return false
+	}
+	headerHeight := f.headerSyncService.Store().Height()
+	dataHeight := f.dataSyncService.Store().Height()
+	if headerHeight > 0 || dataHeight > 0 {
+		return false
+	}
+
+	f.logger.Info().
+		Uint64("store_height", storeHeight).
+		Uint64("header_height", headerHeight).
+		Uint64("data_height", dataHeight).
+		Msg("raft-enabled aggregator with empty sync stores: starting sync services in publisher mode")
+	return true
+}
+
 func (f *failoverState) Run(pCtx context.Context) (multiErr error) {
 	stopService := func(stoppable func(context.Context) error, name string) { //nolint:contextcheck // shutdown uses context.Background intentionally
 		// parent context is cancelled already, so we need to create a new one
@@ -207,15 +237,28 @@ func (f *failoverState) Run(pCtx context.Context) (multiErr error) {
 	})
 
 	// start header and data sync services concurrently to avoid cumulative startup delay.
+	startSyncInPublisherMode := f.shouldStartSyncInPublisherMode(ctx)
 	syncWg, syncCtx := errgroup.WithContext(ctx)
 	syncWg.Go(func() error {
-		if err := f.headerSyncService.Start(syncCtx); err != nil {
+		var err error
+		if startSyncInPublisherMode {
+			err = f.headerSyncService.StartForPublishing(syncCtx)
+		} else {
+			err = f.headerSyncService.Start(syncCtx)
+		}
+		if err != nil {
 			return fmt.Errorf("header sync service: %w", err)
 		}
 		return nil
 	})
 	syncWg.Go(func() error {
-		if err := f.dataSyncService.Start(syncCtx); err != nil {
+		var err error
+		if startSyncInPublisherMode {
+			err = f.dataSyncService.StartForPublishing(syncCtx)
+		} else {
+			err = f.dataSyncService.Start(syncCtx)
+		}
+		if err != nil {
 			return fmt.Errorf("data sync service: %w", err)
 		}
 		return nil

pkg/config/config.go

@@ -400,7 +400,7 @@ type RaftConfig struct {
 	NodeID             string        `mapstructure:"node_id" yaml:"node_id" comment:"Unique identifier for this node in the Raft cluster"`
 	RaftAddr           string        `mapstructure:"raft_addr" yaml:"raft_addr" comment:"Address for Raft communication (host:port)"`
 	RaftDir            string        `mapstructure:"raft_dir" yaml:"raft_dir" comment:"Directory for Raft logs and snapshots"`
-	Bootstrap          bool          `mapstructure:"bootstrap" yaml:"bootstrap" comment:"Bootstrap a new Raft cluster (only for the first node)"`
+	Bootstrap          bool          `mapstructure:"bootstrap" yaml:"bootstrap" comment:"Bootstrap a new static Raft cluster during initial bring-up"`
 	Peers              string        `mapstructure:"peers" yaml:"peers" comment:"Comma-separated list of peer Raft addresses (nodeID@host:port)"`
 	SnapCount          uint64        `mapstructure:"snap_count" yaml:"snap_count" comment:"Number of log entries between snapshots"`
 	SendTimeout        time.Duration `mapstructure:"send_timeout" yaml:"send_timeout" comment:"Max duration to wait for a message to be sent to a peer"`
@@ -646,7 +646,7 @@ func AddFlags(cmd *cobra.Command) {
 	cmd.Flags().String(FlagRaftNodeID, def.Raft.NodeID, "unique identifier for this node in the Raft cluster")
 	cmd.Flags().String(FlagRaftAddr, def.Raft.RaftAddr, "address for Raft communication (host:port)")
 	cmd.Flags().String(FlagRaftDir, def.Raft.RaftDir, "directory for Raft logs and snapshots")
-	cmd.Flags().Bool(FlagRaftBootstrap, def.Raft.Bootstrap, "bootstrap a new Raft cluster (only for the first node)")
+	cmd.Flags().Bool(FlagRaftBootstrap, def.Raft.Bootstrap, "bootstrap a new static Raft cluster during initial bring-up")
 	cmd.Flags().String(FlagRaftPeers, def.Raft.Peers, "comma-separated list of peer Raft addresses (nodeID@host:port)")
 	cmd.Flags().Uint64(FlagRaftSnapCount, def.Raft.SnapCount, "number of log entries between snapshots")
 	cmd.Flags().Duration(FlagRaftSendTimeout, def.Raft.SendTimeout, "max duration to wait for a message to be sent to a peer")