mirrors/tendermint
1
0
Fork 0
mirror of https://github.com/tendermint/tendermint.git synced 2026-08-28 11:56:15 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'feature/abci++vef' into cal/finalize-block

Browse Source
This commit is contained in:
Callum Waters
2022-11-16 14:46:37 +01:00
parent 29b67bb7d2 4d330ab4fc
commit d0b8b8a358
199 changed files with 6654 additions and 3376 deletions
Download Patch File Download Diff File Expand all files Collapse all files
statesync/messages.go
-43
View File
@@ -16,49 +16,6 @@ const (
chunkMsgSize = int(16e6)
)
// mustEncodeMsg encodes a Protobuf message, panicing on error.
func mustEncodeMsg(pb proto.Message) []byte {
msg := ssproto.Message{}
switch pb := pb.(type) {
case *ssproto.ChunkRequest:
msg.Sum = &ssproto.Message_ChunkRequest{ChunkRequest: pb}
case *ssproto.ChunkResponse:
msg.Sum = &ssproto.Message_ChunkResponse{ChunkResponse: pb}
case *ssproto.SnapshotsRequest:
msg.Sum = &ssproto.Message_SnapshotsRequest{SnapshotsRequest: pb}
case *ssproto.SnapshotsResponse:
msg.Sum = &ssproto.Message_SnapshotsResponse{SnapshotsResponse: pb}
default:
panic(fmt.Errorf("unknown message type %T", pb))
}
bz, err := msg.Marshal()
if err != nil {
panic(fmt.Errorf("unable to marshal %T: %w", pb, err))
}
return bz
}
// decodeMsg decodes a Protobuf message.
func decodeMsg(bz []byte) (proto.Message, error) {
pb := &ssproto.Message{}
err := proto.Unmarshal(bz, pb)
if err != nil {
return nil, err
}
switch msg := pb.Sum.(type) {
case *ssproto.Message_ChunkRequest:
return msg.ChunkRequest, nil
case *ssproto.Message_ChunkResponse:
return msg.ChunkResponse, nil
case *ssproto.Message_SnapshotsRequest:
return msg.SnapshotsRequest, nil
case *ssproto.Message_SnapshotsResponse:
return msg.SnapshotsResponse, nil
default:
return nil, fmt.Errorf("unknown message type %T", msg)
}
}
// validateMsg validates a message.
func validateMsg(pb proto.Message) error {
if pb == nil {
statesync/messages_test.go
+4 -2
View File
@@ -7,6 +7,7 @@ import (
"github.com/cosmos/gogoproto/proto"
"github.com/stretchr/testify/require"
"github.com/tendermint/tendermint/p2p"
ssproto "github.com/tendermint/tendermint/proto/tendermint/statesync"
tmproto "github.com/tendermint/tendermint/proto/tendermint/types"
)
@@ -99,8 +100,9 @@ func TestStateSyncVectors(t *testing.T) {
for _, tc := range testCases {
tc := tc
bz := mustEncodeMsg(tc.msg)
w := tc.msg.(p2p.Wrapper).Wrap()
bz, err := proto.Marshal(w)
require.NoError(t, err)
require.Equal(t, tc.expBytes, hex.EncodeToString(bz), tc.testName)
}
statesync/metrics.gen.go
+30
View File
@@ -0,0 +1,30 @@
// Code generated by metricsgen. DO NOT EDIT.
package statesync
import (
"github.com/go-kit/kit/metrics/discard"
prometheus "github.com/go-kit/kit/metrics/prometheus"
stdprometheus "github.com/prometheus/client_golang/prometheus"
)
func PrometheusMetrics(namespace string, labelsAndValues ...string) *Metrics {
labels := []string{}
for i := 0; i < len(labelsAndValues); i += 2 {
labels = append(labels, labelsAndValues[i])
}
return &Metrics{
Syncing: prometheus.NewGaugeFrom(stdprometheus.GaugeOpts{
Namespace: namespace,
Subsystem: MetricsSubsystem,
Name: "syncing",
Help: "Whether or not a node is state syncing. 1 if yes, 0 if no.",
}, labels).With(labelsAndValues...),
}
}
func NopMetrics() *Metrics {
return &Metrics{
Syncing: discard.NewGauge(),
}
}
statesync/metrics.go
+19
View File
@@ -0,0 +1,19 @@
package statesync
import (
"github.com/go-kit/kit/metrics"
)
const (
// MetricsSubsystem is a subsystem shared by all metrics exposed by this
// package.
MetricsSubsystem = "statesync"
)
//go:generate go run ../scripts/metricsgen -struct=Metrics
// Metrics contains metrics exposed by this package.
type Metrics struct {
// Whether or not a node is state syncing. 1 if yes, 0 if no.
Syncing metrics.Gauge
}
statesync/reactor.go
+49 -38
View File
@@ -34,6 +34,7 @@ type Reactor struct {
conn proxy.AppConnSnapshot
connQuery proxy.AppConnQuery
tempDir string
metrics *Metrics
// This will only be set when a state sync is in progress. It is used to feed received
// snapshots and chunks into the sync.
@@ -47,12 +48,14 @@ func NewReactor(
conn proxy.AppConnSnapshot,
connQuery proxy.AppConnQuery,
tempDir string,
metrics *Metrics,
) *Reactor {
r := &Reactor{
cfg: cfg,
conn: conn,
connQuery: connQuery,
metrics: metrics,
}
r.BaseReactor = *p2p.NewBaseReactor("StateSync", r)
@@ -67,12 +70,14 @@ func (r *Reactor) GetChannels() []*p2p.ChannelDescriptor {
Priority: 5,
SendQueueCapacity: 10,
RecvMessageCapacity: snapshotMsgSize,
MessageType: &ssproto.Message{},
},
{
ID: ChunkChannel,
Priority: 3,
SendQueueCapacity: 10,
RecvMessageCapacity: chunkMsgSize,
MessageType: &ssproto.Message{},
},
}
}
@@ -101,27 +106,21 @@ func (r *Reactor) RemovePeer(peer p2p.Peer, reason interface{}) {
}
// Receive implements p2p.Reactor.
func (r *Reactor) Receive(chID byte, src p2p.Peer, msgBytes []byte) {
func (r *Reactor) Receive(e p2p.Envelope) {
if !r.IsRunning() {
return
}
msg, err := decodeMsg(msgBytes)
err := validateMsg(e.Message)
if err != nil {
r.Logger.Error("Error decoding message", "src", src, "chId", chID, "err", err)
r.Switch.StopPeerForError(src, err)
return
}
err = validateMsg(msg)
if err != nil {
r.Logger.Error("Invalid message", "peer", src, "msg", msg, "err", err)
r.Switch.StopPeerForError(src, err)
r.Logger.Error("Invalid message", "peer", e.Src, "msg", e.Message, "err", err)
r.Switch.StopPeerForError(e.Src, err)
return
}
switch chID {
switch e.ChannelID {
case SnapshotChannel:
switch msg := msg.(type) {
switch msg := e.Message.(type) {
case *ssproto.SnapshotsRequest:
snapshots, err := r.recentSnapshots(recentSnapshots)
if err != nil {
@@ -130,14 +129,17 @@ func (r *Reactor) Receive(chID byte, src p2p.Peer, msgBytes []byte) {
}
for _, snapshot := range snapshots {
r.Logger.Debug("Advertising snapshot", "height", snapshot.Height,
"format", snapshot.Format, "peer", src.ID())
src.Send(chID, mustEncodeMsg(&ssproto.SnapshotsResponse{
Height: snapshot.Height,
Format: snapshot.Format,
Chunks: snapshot.Chunks,
Hash: snapshot.Hash,
Metadata: snapshot.Metadata,
}))
"format", snapshot.Format, "peer", e.Src.ID())
e.Src.Send(p2p.Envelope{
ChannelID: e.ChannelID,
Message: &ssproto.SnapshotsResponse{
Height: snapshot.Height,
Format: snapshot.Format,
Chunks: snapshot.Chunks,
Hash: snapshot.Hash,
Metadata: snapshot.Metadata,
},
})
}
case *ssproto.SnapshotsResponse:
@@ -147,8 +149,8 @@ func (r *Reactor) Receive(chID byte, src p2p.Peer, msgBytes []byte) {
r.Logger.Debug("Received unexpected snapshot, no state sync in progress")
return
}
r.Logger.Debug("Received snapshot", "height", msg.Height, "format", msg.Format, "peer", src.ID())
_, err := r.syncer.AddSnapshot(src, &snapshot{
r.Logger.Debug("Received snapshot", "height", msg.Height, "format", msg.Format, "peer", e.Src.ID())
_, err := r.syncer.AddSnapshot(e.Src, &snapshot{
Height: msg.Height,
Format: msg.Format,
Chunks: msg.Chunks,
@@ -158,7 +160,7 @@ func (r *Reactor) Receive(chID byte, src p2p.Peer, msgBytes []byte) {
// TODO: We may want to consider punishing the peer for certain errors
if err != nil {
r.Logger.Error("Failed to add snapshot", "height", msg.Height, "format", msg.Format,
"peer", src.ID(), "err", err)
"peer", e.Src.ID(), "err", err)
return
}
@@ -167,10 +169,10 @@ func (r *Reactor) Receive(chID byte, src p2p.Peer, msgBytes []byte) {
}
case ChunkChannel:
switch msg := msg.(type) {
switch msg := e.Message.(type) {
case *ssproto.ChunkRequest:
r.Logger.Debug("Received chunk request", "height", msg.Height, "format", msg.Format,
"chunk", msg.Index, "peer", src.ID())
"chunk", msg.Index, "peer", e.Src.ID())
resp, err := r.conn.LoadSnapshotChunk(context.TODO(), &abci.RequestLoadSnapshotChunk{
Height: msg.Height,
Format: msg.Format,
@@ -182,30 +184,33 @@ func (r *Reactor) Receive(chID byte, src p2p.Peer, msgBytes []byte) {
return
}
r.Logger.Debug("Sending chunk", "height", msg.Height, "format", msg.Format,
"chunk", msg.Index, "peer", src.ID())
src.Send(ChunkChannel, mustEncodeMsg(&ssproto.ChunkResponse{
Height: msg.Height,
Format: msg.Format,
Index: msg.Index,
Chunk: resp.Chunk,
Missing: resp.Chunk == nil,
}))
"chunk", msg.Index, "peer", e.Src.ID())
e.Src.Send(p2p.Envelope{
ChannelID: ChunkChannel,
Message: &ssproto.ChunkResponse{
Height: msg.Height,
Format: msg.Format,
Index: msg.Index,
Chunk: resp.Chunk,
Missing: resp.Chunk == nil,
},
})
case *ssproto.ChunkResponse:
r.mtx.RLock()
defer r.mtx.RUnlock()
if r.syncer == nil {
r.Logger.Debug("Received unexpected chunk, no state sync in progress", "peer", src.ID())
r.Logger.Debug("Received unexpected chunk, no state sync in progress", "peer", e.Src.ID())
return
}
r.Logger.Debug("Received chunk, adding to sync", "height", msg.Height, "format", msg.Format,
"chunk", msg.Index, "peer", src.ID())
"chunk", msg.Index, "peer", e.Src.ID())
_, err := r.syncer.AddChunk(&chunk{
Height: msg.Height,
Format: msg.Format,
Index: msg.Index,
Chunk: msg.Chunk,
Sender: src.ID(),
Sender: e.Src.ID(),
})
if err != nil {
r.Logger.Error("Failed to add chunk", "height", msg.Height, "format", msg.Format,
@@ -218,7 +223,7 @@ func (r *Reactor) Receive(chID byte, src p2p.Peer, msgBytes []byte) {
}
default:
r.Logger.Error("Received message on invalid channel %x", chID)
r.Logger.Error("Received message on invalid channel %x", e.ChannelID)
}
}
@@ -264,13 +269,18 @@ func (r *Reactor) Sync(stateProvider StateProvider, discoveryTime time.Duration)
r.mtx.Unlock()
return sm.State{}, nil, errors.New("a state sync is already in progress")
}
r.metrics.Syncing.Set(1)
r.syncer = newSyncer(r.cfg, r.Logger, r.conn, r.connQuery, stateProvider, r.tempDir)
r.mtx.Unlock()
hook := func() {
r.Logger.Debug("Requesting snapshots from known peers")
// Request snapshots from all currently connected peers
r.Switch.Broadcast(SnapshotChannel, mustEncodeMsg(&ssproto.SnapshotsRequest{}))
r.Switch.Broadcast(p2p.Envelope{
ChannelID: SnapshotChannel,
Message: &ssproto.SnapshotsRequest{},
})
}
hook()
@@ -279,6 +289,7 @@ func (r *Reactor) Sync(stateProvider StateProvider, discoveryTime time.Duration)
r.mtx.Lock()
r.syncer = nil
r.metrics.Syncing.Set(0)
r.mtx.Unlock()
return state, commit, err
}
statesync/reactor_test.go
+35 -10
View File
@@ -4,6 +4,7 @@ import (
"testing"
"time"
"github.com/cosmos/gogoproto/proto"
"github.com/stretchr/testify/assert"
"github.com/stretchr/testify/mock"
"github.com/stretchr/testify/require"
@@ -53,16 +54,24 @@ func TestReactor_Receive_ChunkRequest(t *testing.T) {
peer.On("ID").Return(p2p.ID("id"))
var response *ssproto.ChunkResponse
if tc.expectResponse != nil {
peer.On("Send", ChunkChannel, mock.Anything).Run(func(args mock.Arguments) {
msg, err := decodeMsg(args[1].([]byte))
peer.On("Send", mock.MatchedBy(func(i interface{}) bool {
e, ok := i.(p2p.Envelope)
return ok && e.ChannelID == ChunkChannel
})).Run(func(args mock.Arguments) {
e := args[0].(p2p.Envelope)
// Marshal to simulate a wire roundtrip.
bz, err := proto.Marshal(e.Message)
require.NoError(t, err)
response = msg.(*ssproto.ChunkResponse)
err = proto.Unmarshal(bz, e.Message)
require.NoError(t, err)
response = e.Message.(*ssproto.ChunkResponse)
}).Return(true)
}
// Start a reactor and send a ssproto.ChunkRequest, then wait for and check response
cfg := config.DefaultStateSyncConfig()
r := NewReactor(*cfg, conn, nil, "")
r := NewReactor(*cfg, conn, nil, "", NopMetrics())
err := r.Start()
require.NoError(t, err)
t.Cleanup(func() {
@@ -71,7 +80,11 @@ func TestReactor_Receive_ChunkRequest(t *testing.T) {
}
})
r.Receive(ChunkChannel, peer, mustEncodeMsg(tc.request))
r.Receive(p2p.Envelope{
ChannelID: ChunkChannel,
Src: peer,
Message: tc.request,
})
time.Sleep(100 * time.Millisecond)
assert.Equal(t, tc.expectResponse, response)
@@ -131,16 +144,24 @@ func TestReactor_Receive_SnapshotsRequest(t *testing.T) {
peer := &p2pmocks.Peer{}
if len(tc.expectResponses) > 0 {
peer.On("ID").Return(p2p.ID("id"))
peer.On("Send", SnapshotChannel, mock.Anything).Run(func(args mock.Arguments) {
msg, err := decodeMsg(args[1].([]byte))
peer.On("Send", mock.MatchedBy(func(i interface{}) bool {
e, ok := i.(p2p.Envelope)
return ok && e.ChannelID == SnapshotChannel
})).Run(func(args mock.Arguments) {
e := args[0].(p2p.Envelope)
// Marshal to simulate a wire roundtrip.
bz, err := proto.Marshal(e.Message)
require.NoError(t, err)
responses = append(responses, msg.(*ssproto.SnapshotsResponse))
err = proto.Unmarshal(bz, e.Message)
require.NoError(t, err)
responses = append(responses, e.Message.(*ssproto.SnapshotsResponse))
}).Return(true)
}
// Start a reactor and send a SnapshotsRequestMessage, then wait for and check responses
cfg := config.DefaultStateSyncConfig()
r := NewReactor(*cfg, conn, nil, "")
r := NewReactor(*cfg, conn, nil, "", NopMetrics())
err := r.Start()
require.NoError(t, err)
t.Cleanup(func() {
@@ -149,7 +170,11 @@ func TestReactor_Receive_SnapshotsRequest(t *testing.T) {
}
})
r.Receive(SnapshotChannel, peer, mustEncodeMsg(&ssproto.SnapshotsRequest{}))
r.Receive(p2p.Envelope{
ChannelID: SnapshotChannel,
Src: peer,
Message: &ssproto.SnapshotsRequest{},
})
time.Sleep(100 * time.Millisecond)
assert.Equal(t, tc.expectResponses, responses)
statesync/syncer.go
+13 -6
View File
@@ -126,7 +126,11 @@ func (s *syncer) AddSnapshot(peer p2p.Peer, snapshot *snapshot) (bool, error) {
// to discover snapshots, later we may want to do retries and stuff.
func (s *syncer) AddPeer(peer p2p.Peer) {
s.logger.Debug("Requesting snapshots from peer", "peer", peer.ID())
peer.Send(SnapshotChannel, mustEncodeMsg(&ssproto.SnapshotsRequest{}))
e := p2p.Envelope{
ChannelID: SnapshotChannel,
Message: &ssproto.SnapshotsRequest{},
}
peer.Send(e)
}
// RemovePeer removes a peer from the pool.
@@ -467,11 +471,14 @@ func (s *syncer) requestChunk(snapshot *snapshot, chunk uint32) {
}
s.logger.Debug("Requesting snapshot chunk", "height", snapshot.Height,
"format", snapshot.Format, "chunk", chunk, "peer", peer.ID())
peer.Send(ChunkChannel, mustEncodeMsg(&ssproto.ChunkRequest{
Height: snapshot.Height,
Format: snapshot.Format,
Index: chunk,
}))
peer.Send(p2p.Envelope{
ChannelID: ChunkChannel,
Message: &ssproto.ChunkRequest{
Height: snapshot.Height,
Format: snapshot.Format,
Index: chunk,
},
})
}
// verifyApp verifies the sync, checking the app hash, last block height and app version
statesync/syncer_test.go
+27 -7
View File
@@ -98,13 +98,27 @@ func TestSyncer_SyncAny(t *testing.T) {
// Adding a couple of peers should trigger snapshot discovery messages
peerA := &p2pmocks.Peer{}
peerA.On("ID").Return(p2p.ID("a"))
peerA.On("Send", SnapshotChannel, mustEncodeMsg(&ssproto.SnapshotsRequest{})).Return(true)
peerA.On("Send", mock.MatchedBy(func(i interface{}) bool {
e, ok := i.(p2p.Envelope)
if !ok {
return false
}
req, ok := e.Message.(*ssproto.SnapshotsRequest)
return ok && e.ChannelID == SnapshotChannel && req != nil
})).Return(true)
syncer.AddPeer(peerA)
peerA.AssertExpectations(t)
peerB := &p2pmocks.Peer{}
peerB.On("ID").Return(p2p.ID("b"))
peerB.On("Send", SnapshotChannel, mustEncodeMsg(&ssproto.SnapshotsRequest{})).Return(true)
peerB.On("Send", mock.MatchedBy(func(i interface{}) bool {
e, ok := i.(p2p.Envelope)
if !ok {
return false
}
req, ok := e.Message.(*ssproto.SnapshotsRequest)
return ok && e.ChannelID == SnapshotChannel && req != nil
})).Return(true)
syncer.AddPeer(peerB)
peerB.AssertExpectations(t)
@@ -147,9 +161,9 @@ func TestSyncer_SyncAny(t *testing.T) {
chunkRequests := make(map[uint32]int)
chunkRequestsMtx := tmsync.Mutex{}
onChunkRequest := func(args mock.Arguments) {
pb, err := decodeMsg(args[1].([]byte))
require.NoError(t, err)
msg := pb.(*ssproto.ChunkRequest)
e, ok := args[0].(p2p.Envelope)
require.True(t, ok)
msg := e.Message.(*ssproto.ChunkRequest)
require.EqualValues(t, 1, msg.Height)
require.EqualValues(t, 1, msg.Format)
require.LessOrEqual(t, msg.Index, uint32(len(chunks)))
@@ -162,8 +176,14 @@ func TestSyncer_SyncAny(t *testing.T) {
chunkRequests[msg.Index]++
chunkRequestsMtx.Unlock()
}
peerA.On("Send", ChunkChannel, mock.Anything).Maybe().Run(onChunkRequest).Return(true)
peerB.On("Send", ChunkChannel, mock.Anything).Maybe().Run(onChunkRequest).Return(true)
peerA.On("Send", mock.MatchedBy(func(i interface{}) bool {
e, ok := i.(p2p.Envelope)
return ok && e.ChannelID == ChunkChannel
})).Maybe().Run(onChunkRequest).Return(true)
peerB.On("Send", mock.MatchedBy(func(i interface{}) bool {
e, ok := i.(p2p.Envelope)
return ok && e.ChannelID == ChunkChannel
})).Maybe().Run(onChunkRequest).Return(true)
// The first time we're applying chunk 2 we tell it to retry the snapshot and discard chunk 1,
// which should cause it to keep the existing chunk 0 and 2, and restart restoration from