support repair snapshot

app/repair_state.go

@@ -89,7 +89,8 @@ func repairStateOnStart(ctx *server.Context) {
 func RepairState(ctx *server.Context, onStart bool) {
 	sm.SetIgnoreSmbCheck(true)
 	iavl.SetIgnoreVersionCheck(true)
-	rootmulti.SetRepair()
+	global.SetRepairState(true)
+	defer global.SetRepairState(false)
 
 	// load latest block height
 	dataDir := filepath.Join(ctx.Config.RootDir, "data")

cmd/okbchaind/repair_data.go

@@ -2,7 +2,6 @@ package main
 
 import (
 	"fmt"
-	"github.com/okx/okbchain/libs/cosmos-sdk/store/mpt"
 	"log"
 	"net/http"
 	_ "net/http/pprof"
@@ -64,7 +63,4 @@ func setExternalPackageValue() {
 		tmiavl.SetEnableFastStorage(true)
 		tmiavl.SetIgnoreAutoUpgrade(true)
 	}
-	if !viper.GetBool(tmiavl.FlagIavlDiscardFastStorage) {
-		mpt.SetSnapshotRebuild(true)
-	}
 }

libs/cosmos-sdk/store/mpt/snapshot.go

@@ -6,6 +6,7 @@ import (
 	"github.com/ethereum/go-ethereum/core/rawdb"
 	"github.com/ethereum/go-ethereum/core/state/snapshot"
 	mpttypes "github.com/okx/okbchain/libs/cosmos-sdk/store/mpt/types"
+	"github.com/okx/okbchain/libs/tendermint/global"
 )
 
 var (
@@ -15,8 +16,10 @@ var (
 
 const (
 	// snapshotMemoryLayerCount snapshot memory layer count
-	// as we dont rollback transactions so we only keep 1 memory layer
-	snapshotMemoryLayerCount = 1
+	// snapshotMemoryLayerCount controls the snapshot Journal height,
+	// if repair start-height is lower than snapshot Journal height,
+	// snapshot will not be repaired anymore
+	snapshotMemoryLayerCount = 10
 )
 
 func DisableSnapshot() {
@@ -42,6 +45,9 @@ func (ms *MptStore) openSnapshot() error {
 		ms.logger.Error("Enabling snapshot recovery", "chainhead", version, "diskbase", *layer)
 		recovery = true
 	}
+	if global.GetRepairState() {
+		recovery = true
+	}
 	var err error
 	ms.snaps, err = snapshot.NewCustom(ms.db.TrieDB().DiskDB(), ms.db.TrieDB(), 256, ms.originalRoot, false, gSnapshotRebuild, recovery, ms.retriever)
 	if err != nil {
@@ -74,6 +80,8 @@ func (ms *MptStore) prepareSnap(root common.Hash) {
 		ms.snapDestructs = make(map[common.Hash]struct{})
 		ms.snapAccounts = make(map[common.Hash][]byte)
 		ms.snapStorage = make(map[common.Hash]map[common.Hash][]byte)
+	} else {
+		ms.logger.Error("prepare snapshot error", "root", root)
 	}
 }
 
@@ -94,6 +102,9 @@ func (ms *MptStore) commitSnap(root common.Hash) {
 		if err := ms.snaps.Cap(root, snapshotMemoryLayerCount); err != nil {
 			ms.logger.Error("Failed to cap snapshot tree", "root", root, "layers", snapshotMemoryLayerCount, "err", err)
 		}
+		if _, err := ms.snaps.Journal(root); err != nil {
+			ms.logger.Error("Failed to journal snapshot tree", "root", root, "err", err)
+		}
 	}
 	ms.snap, ms.snapDestructs, ms.snapAccounts, ms.snapStorage = nil, nil, nil, nil
 

libs/cosmos-sdk/store/rootmulti/rootmulti_store.go

@@ -3,6 +3,7 @@ package rootmulti
 import (
 	"encoding/binary"
 	"fmt"
+	"github.com/okx/okbchain/libs/tendermint/global"
 	"io"
 	"log"
 	"path/filepath"
@@ -47,18 +48,6 @@ const (
 	maxPruneHeightsLength = 100
 )
 
-var (
-	repairing bool
-)
-
-func SetRepair() {
-	repairing = true
-}
-
-func getRepair() bool {
-	return repairing
-}
-
 // Store is composed of many CommitStores. Name contrasts with
 // cacheMultiStore which is for cache-wrapping other MultiStores. It implements
 // the CommitMultiStore interface.
@@ -408,7 +397,7 @@ func (rs *Store) loadVersion(ver int64, upgrades *types.StoreUpgrades) error {
 						// we can not get the upgrade version before the expect height,
 						// and we should not use the original 0 too, because 0 means the latest height,
 						// so when we repair data before the milestone. we open a empty tree by cur version.
-						if getRepair() && version == 0 {
+						if global.GetRepairState() && version == 0 {
 							param.upgradeVersion = uint64(ver)
 						}
 						rs.storesParams[key] = param

libs/tendermint/global/status.go

@@ -0,0 +1,11 @@
+package global
+
+var repairState bool
+
+func SetRepairState(state bool) {
+	repairState = state
+}
+
+func GetRepairState() bool {
+	return repairState
+}