State sync support

CHANGELOG.md

@@ -81,6 +81,14 @@ to now accept a `codec.JSONMarshaler` for modular serialization of genesis state
 * (keys) [\#5820](https://github.com/cosmos/cosmos-sdk/pull/5820/) Removed method CloseDB from Keybase interface.
 * (baseapp) [\#5837](https://github.com/cosmos/cosmos-sdk/issues/5837) Transaction simulation now returns a `SimulationResponse` which contains the `GasInfo` and
 `Result` from the execution.
+* (store) [\#5803](https://github.com/cosmos/cosmos-sdk/pull/5803) The `store.CommitMultiStore` interface now includes the new `snapshots.Snapshotter` interface as well.
+* (crypto/keyring) [\#5866](https://github.com/cosmos/cosmos-sdk/pull/5866) Move `Keyring` and `Keybase` implementations and their associated types from `crypto/keys/` to `crypto/keyring/`.
+* (crypto) [\#5880](https://github.com/cosmos/cosmos-sdk/pull/5880) Merge `crypto/keys/mintkey` into `crypto`.
+* (crypto/keyring) [\#5858](https://github.com/cosmos/cosmos-sdk/pull/5858) Make Keyring store keys by name and address's hexbytes representation.
+* (crypto/keyring) [\#5889](https://github.com/cosmos/cosmos-sdk/pull/5889) Deprecate old keybase implementation:
+  - Remove `Update` from the `Keybase` interface.
+  - `NewKeyring()` now accepts a new backend: `MemoryBackend`.
+  - `New()` has been renamed to`NewLegacy()`, which now returns a `LegacyKeybase` type that only allows migration of keys from the legacy keybase to a new keyring.
 * (client/input) [\#5904](https://github.com/cosmos/cosmos-sdk/pull/5904) Removal of unnecessary `GetCheckPassword`, `PrintPrefixed` functions.
 * (client/keys) [\#5889](https://github.com/cosmos/cosmos-sdk/pull/5889) Rename `NewKeyBaseFromDir()` -> `NewLegacyKeyBaseFromDir()`.
 * (crypto) [\#5880](https://github.com/cosmos/cosmos-sdk/pull/5880) Merge `crypto/keys/mintkey` into `crypto`.
@@ -99,6 +107,8 @@ information on how to implement the new `Keyring` interface.
 ### Features
 
 * (x/ibc) [\#5588](https://github.com/cosmos/cosmos-sdk/pull/5588) Add [ICS 024 - Host State Machine Requirements](https://github.com/cosmos/ics/tree/master/spec/ics-024-host-requirements) subpackage to `x/ibc` module.
+* (baseapp) [\#5803](https://github.com/cosmos/cosmos-sdk/pull/5803) Added support for taking state snapshots at regular height intervals, via options `snapshot-interval` and `snapshot-retention`.
+* (store) [\#5803](https://github.com/cosmos/cosmos-sdk/pull/5803) Added `rootmulti.Store` methods for taking and restoring snapshots, based on `iavl.Store` export/import.
 * (x/ibc) [\#5277](https://github.com/cosmos/cosmos-sdk/pull/5277) `x/ibc` changes from IBC alpha. For more details check the the  [`x/ibc/spec`](https://github.com/cosmos/tree/master/x/ibc/spec) directory:
   * [ICS 002 - Client Semantics](https://github.com/cosmos/ics/tree/master/spec/ics-002-client-semantics) subpackage
   * [ICS 003 - Connection Semantics](https://github.com/cosmos/ics/blob/master/spec/ics-003-connection-semantics) subpackage

baseapp/abci.go

@@ -1,6 +1,7 @@
 package baseapp
 
 import (
+	"errors"
 	"fmt"
 	"os"
 	"sort"
@@ -10,6 +11,7 @@ import (
 	abci "github.com/tendermint/tendermint/abci/types"
 
 	"github.com/cosmos/cosmos-sdk/codec"
+	"github.com/cosmos/cosmos-sdk/snapshots"
 	sdk "github.com/cosmos/cosmos-sdk/types"
 	sdkerrors "github.com/cosmos/cosmos-sdk/types/errors"
 )
@@ -263,6 +265,10 @@ func (app *BaseApp) Commit() (res abci.ResponseCommit) {
 		app.halt()
 	}
 
+	if app.snapshotInterval > 0 && uint64(header.Height)%app.snapshotInterval == 0 {
+		go app.snapshot(header.Height)
+	}
+
 	return abci.ResponseCommit{
 		Data: commitID.Hash,
 	}
@@ -290,6 +296,27 @@ func (app *BaseApp) halt() {
 	os.Exit(0)
 }
 
+// snapshot takes a snapshot of the current state and prunes any old snapshots.
+func (app *BaseApp) snapshot(height int64) {
+	app.logger.Info("Taking state snapshot", "height", height)
+	snapshot, err := app.snapshotManager.Take(uint64(height))
+	if err != nil {
+		app.logger.Error("Failed to take state snapshot", "height", height, "err", err)
+		return
+	}
+	app.logger.Info("Completed state snapshot", "height", height, "format", snapshot.Format)
+
+	if snapshotRetention > 0 {
+		app.logger.Debug("Pruning state snapshots")
+		pruned, err := app.snapshotManager.Prune(snapshotRetention)
+		if err != nil {
+			app.logger.Error("Failed to prune state snapshots", "err", err.Error())
+			return
+		}
+		app.logger.Debug("Pruned state snapshots", "pruned", pruned)
+	}
+}
+
 // Query implements the ABCI interface. It delegates to CommitMultiStore if it
 // implements Queryable.
 func (app *BaseApp) Query(req abci.RequestQuery) abci.ResponseQuery {
@@ -316,6 +343,100 @@ func (app *BaseApp) Query(req abci.RequestQuery) abci.ResponseQuery {
 	return sdkerrors.QueryResult(sdkerrors.Wrap(sdkerrors.ErrUnknownRequest, "unknown query path"))
 }
 
+// ListSnapshots implements the ABCI interface. It delegates to app.snapshotManager if set.
+func (app *BaseApp) ListSnapshots(req abci.RequestListSnapshots) abci.ResponseListSnapshots {
+	resp := abci.ResponseListSnapshots{Snapshots: []*abci.Snapshot{}}
+	if app.snapshotManager == nil {
+		return resp
+	}
+
+	snapshots, err := app.snapshotManager.List()
+	if err != nil {
+		app.logger.Error("Failed to list snapshots", "err", err.Error())
+		return resp
+	}
+	for _, snapshot := range snapshots {
+		abciSnapshot, err := snapshot.ToABCI()
+		if err != nil {
+			app.logger.Error("Failed to list snapshots", "err", err.Error())
+			return resp
+		}
+		resp.Snapshots = append(resp.Snapshots, &abciSnapshot)
+	}
+
+	return resp
+}
+
+// LoadSnapshotChunk implements the ABCI interface. It delegates to app.snapshotManager if set.
+func (app *BaseApp) LoadSnapshotChunk(req abci.RequestLoadSnapshotChunk) abci.ResponseLoadSnapshotChunk {
+	if app.snapshotManager == nil {
+		return abci.ResponseLoadSnapshotChunk{}
+	}
+	chunk, err := app.snapshotManager.LoadChunk(req.Height, req.Format, req.Chunk)
+	if err != nil {
+		app.logger.Error("Failed to load snapshot chunk", "height", req.Height, "format", req.Format,
+			"chunk", req.Chunk, "err", err.Error())
+		return abci.ResponseLoadSnapshotChunk{}
+	}
+	return abci.ResponseLoadSnapshotChunk{Chunk: chunk}
+}
+
+// OfferSnapshot implements the ABCI interface. It delegates to app.snapshotManager if set.
+func (app *BaseApp) OfferSnapshot(req abci.RequestOfferSnapshot) abci.ResponseOfferSnapshot {
+	if req.Snapshot == nil {
+		app.logger.Error("Received nil snapshot")
+		return abci.ResponseOfferSnapshot{Result: abci.ResponseOfferSnapshot_REJECT}
+	}
+
+	snapshot, err := snapshots.SnapshotFromABCI(req.Snapshot)
+	if err != nil {
+		app.logger.Error("Failed to decode snapshot metadata", "err", err)
+		return abci.ResponseOfferSnapshot{Result: abci.ResponseOfferSnapshot_REJECT}
+	}
+	err = app.snapshotManager.Restore(snapshot)
+	switch {
+	case err == nil:
+		return abci.ResponseOfferSnapshot{Result: abci.ResponseOfferSnapshot_ACCEPT}
+
+	case errors.Is(err, snapshots.ErrUnknownFormat):
+		return abci.ResponseOfferSnapshot{Result: abci.ResponseOfferSnapshot_REJECT_FORMAT}
+
+	case errors.Is(err, snapshots.ErrInvalidMetadata):
+		app.logger.Error("Rejecting invalid snapshot", "height", req.Snapshot.Height,
+			"format", req.Snapshot.Format, "err", err.Error())
+		return abci.ResponseOfferSnapshot{Result: abci.ResponseOfferSnapshot_REJECT}
+
+	default:
+		app.logger.Error("Failed to restore snapshot", "height", req.Snapshot.Height,
+			"format", req.Snapshot.Format, "err", err.Error())
+		// We currently don't support resetting the IAVL stores and retrying a different snapshot,
+		// so we ask Tendermint to abort all snapshot restoration.
+		return abci.ResponseOfferSnapshot{Result: abci.ResponseOfferSnapshot_ABORT}
+	}
+}
+
+// ApplySnapshotChunk implements the ABCI interface. It delegates to app.snapshotManager if set.
+func (app *BaseApp) ApplySnapshotChunk(req abci.RequestApplySnapshotChunk) abci.ResponseApplySnapshotChunk {
+	_, err := app.snapshotManager.RestoreChunk(req.Chunk)
+	switch {
+	case err == nil:
+		return abci.ResponseApplySnapshotChunk{Result: abci.ResponseApplySnapshotChunk_ACCEPT}
+
+	case errors.Is(err, snapshots.ErrChunkHashMismatch):
+		app.logger.Error("Chunk checksum mismatch, rejecting sender and requesting refetch",
+			"chunk", req.Index, "sender", req.Sender, "err", err)
+		return abci.ResponseApplySnapshotChunk{
+			Result:        abci.ResponseApplySnapshotChunk_RETRY,
+			RefetchChunks: []uint32{req.Index},
+			RejectSenders: []string{req.Sender},
+		}
+
+	default:
+		app.logger.Error("Failed to restore snapshot", "err", err.Error())
+		return abci.ResponseApplySnapshotChunk{Result: abci.ResponseApplySnapshotChunk_ABORT}
+	}
+}
+
 func handleQueryApp(app *BaseApp, path []string, req abci.RequestQuery) abci.ResponseQuery {
 	if len(path) >= 2 {
 		switch path[1] {

baseapp/baseapp.go

@@ -11,12 +11,14 @@ import (
 	"github.com/tendermint/tendermint/libs/log"
 	dbm "github.com/tendermint/tm-db"
 
+	"github.com/cosmos/cosmos-sdk/snapshots"
 	"github.com/cosmos/cosmos-sdk/store"
 	sdk "github.com/cosmos/cosmos-sdk/types"
 	sdkerrors "github.com/cosmos/cosmos-sdk/types/errors"
 )
 
 const (
+	snapshotRetention           = 2
 	runTxModeCheck    runTxMode = iota // Check a transaction
 	runTxModeReCheck                   // Recheck a (pending) transaction after a commit
 	runTxModeSimulate                  // Simulate a transaction
@@ -62,6 +64,10 @@ type BaseApp struct { // nolint: maligned
 	idPeerFilter   sdk.PeerFilter   // filter peers by node ID
 	fauxMerkleMode bool             // if true, IAVL MountStores uses MountStoresDB for simulation speed.
 
+	// manages snapshots, i.e. dumps of app state at certain intervals
+	snapshotManager  *snapshots.Manager
+	snapshotInterval uint64 // block interval between snapshots - must be multiple of IAVL SnapshotEvery
+
 	// volatile states:
 	//
 	// checkState is set on InitChain and reset on Commit