Prepare changelog for Release v0.35.8 (#8988)

Update the queue documentation to reflect the types of queues and current default queue.

(cherry picked from commit c1c501ecd4)

Co-authored-by: William Banfield <4561443+williambanfield@users.noreply.github.com>

CHANGELOG.md

@@ -2,6 +2,41 @@
 
 Friendly reminder: We have a [bug bounty program](https://hackerone.com/cosmos).
 
+## v0.35.8
+
+July 12, 2022
+
+Special thanks to external contributors on this release: @joeabbey
+
+This release fixes an unbounded heap growth issue in the implementation of the
+priority mempool, as well as some configuration, logging, and peer dialing
+improvements in the non-legacy p2p stack. It also adds a new opt-in
+"simple-priority" value for the `p2p.queue-type` setting, that should improve
+gossip performance for non-legacy peer networks.
+
+### BREAKING CHANGES
+
+- CLI/RPC/Config
+
+   - [node] [\#8902](https://github.com/tendermint/tendermint/pull/8902) Always start blocksync and avoid misconfiguration (@tychoish)
+
+### FEATURES
+
+- [cli] [\#8675](https://github.com/tendermint/tendermint/pull/8675) Add command to force compact goleveldb databases (@cmwaters)
+
+### IMPROVEMENTS
+
+- [p2p] [\#8914](https://github.com/tendermint/tendermint/pull/8914) [\#8875](https://github.com/tendermint/tendermint/pull/8875) Improvements to peer dialing (backported). (@tychoish)
+- [p2p] [\#8820](https://github.com/tendermint/tendermint/pull/8820) add eviction metrics and cleanup dialing error handling (backport #8819) (@tychoish)
+- [logging] [\#8896](https://github.com/tendermint/tendermint/pull/8896) Do not pre-process log results (backport #8895). (@tychoish)
+- [p2p] [\#8956](https://github.com/tendermint/tendermint/pull/8956) Simpler priority queue (backport #8929). (@tychoish)
+
+### BUG FIXES
+
+- [mempool] [\#8944](https://github.com/tendermint/tendermint/pull/8944) Fix unbounded heap growth in the priority mempool. (@creachadair)
+- [p2p] [\#8869](https://github.com/tendermint/tendermint/pull/8869) Set empty timeouts to configed values. (backport #8847). (@williambanfield)
+
+
 ## v0.35.7
 
 June 16, 2022

CHANGELOG_PENDING.md

@@ -2,7 +2,7 @@
 
 Friendly reminder: We have a [bug bounty program](https://hackerone.com/cosmos).
 
-## v0.35.8
+## v0.35.9
 
 Month DD, YYYY
 
@@ -22,10 +22,6 @@ Special thanks to external contributors on this release:
 
 ### FEATURES
 
-- [cli] [\#8675] Add command to force compact goleveldb databases (@cmwaters)
-
 ### IMPROVEMENTS
 
 ### BUG FIXES
-
-- [mempool] \#8944 Fix unbounded heap growth in the priority mempool. (@creachadair)

cmd/tendermint/commands/run_node.go

@@ -34,9 +34,6 @@ func AddNodeFlags(cmd *cobra.Command) {
 		config.PrivValidator.ListenAddr,
 		"socket address to listen on for connections from external priv-validator process")
 
-	// node flags
-	cmd.Flags().Bool("blocksync.enable", config.BlockSync.Enable, "enable fast blockchain syncing")
-
 	// TODO (https://github.com/tendermint/tendermint/issues/6908): remove this check after the v0.35 release cycle
 	// This check was added to give users an upgrade prompt to use the new flag for syncing.
 	//

config/config.go

@@ -762,7 +762,7 @@ type P2PConfig struct { //nolint: maligned
 	UseLegacy bool `mapstructure:"use-legacy"`
 
 	// Makes it possible to configure which queue backend the p2p
-	// layer uses. Options are: "fifo", "priority" and "wdrr",
+	// layer uses. Options are: "fifo", "simple-priority", "priority", and "wdrr",
 	// with the default being "priority".
 	QueueType string `mapstructure:"queue-type"`
 }

config/toml.go

@@ -298,10 +298,11 @@ pprof-laddr = "{{ .RPC.PprofListenAddress }}"
 #######################################################
 [p2p]
 
-# Enable the legacy p2p layer.
+# Select the p2p internal queue.
+# Options are: "fifo", "simple-priority", "priority", and "wdrr"
+# with the default being "priority".
 use-legacy = {{ .P2P.UseLegacy }}
 
-# Select the p2p internal queue
 queue-type = "{{ .P2P.QueueType }}"
 
 # Address to listen for incoming connections

docs/introduction/quick-start.md

@@ -106,10 +106,10 @@ Next, use the `tendermint testnet` command to create four directories of config
 Before you can start the network, you'll need peers identifiers (IPs are not enough and can change). We'll refer to them as ID1, ID2, ID3, ID4.
 
 ```sh
-tendermint show_node_id --home ./mytestnet/node0
-tendermint show_node_id --home ./mytestnet/node1
-tendermint show_node_id --home ./mytestnet/node2
-tendermint show_node_id --home ./mytestnet/node3
+tendermint show-node-id --home ./mytestnet/node0
+tendermint show-node-id --home ./mytestnet/node1
+tendermint show-node-id --home ./mytestnet/node2
+tendermint show-node-id --home ./mytestnet/node3
 ```
 
 Finally, from each machine, run:

internal/mempool/cache.go

@@ -22,6 +22,10 @@ type TxCache interface {
 
 	// Remove removes the given raw transaction from the cache.
 	Remove(tx types.Tx)
+
+	// Has reports whether tx is present in the cache. Checking for presence is
+	// not treated as an access of the value.
+	Has(tx types.Tx) bool
 }
 
 var _ TxCache = (*LRUTxCache)(nil)
@@ -97,6 +101,14 @@ func (c *LRUTxCache) Remove(tx types.Tx) {
 	}
 }
 
+func (c *LRUTxCache) Has(tx types.Tx) bool {
+	c.mtx.Lock()
+	defer c.mtx.Unlock()
+
+	_, ok := c.cacheMap[tx.Key()]
+	return ok
+}
+
 // NopTxCache defines a no-op raw transaction cache.
 type NopTxCache struct{}
 
@@ -105,3 +117,4 @@ var _ TxCache = (*NopTxCache)(nil)
 func (NopTxCache) Reset()             {}
 func (NopTxCache) Push(types.Tx) bool { return true }
 func (NopTxCache) Remove(types.Tx)    {}
+func (NopTxCache) Has(types.Tx) bool  { return false }

internal/mempool/v1/mempool.go

@@ -597,6 +597,7 @@ func (txmp *TxMempool) initialTxCallback(wtx *WrappedTx, res *abci.Response) {
 				"old_priority", w.priority,
 			)
 			txmp.removeTxByElement(vic)
+			txmp.cache.Remove(w.tx)
 			txmp.metrics.EvictedTxs.Add(1)
 
 			// We may not need to evict all the eligible transactions.  Bail out
@@ -783,9 +784,11 @@ func (txmp *TxMempool) purgeExpiredTxs(blockHeight int64) {
 		w := cur.Value.(*WrappedTx)
 		if txmp.config.TTLNumBlocks > 0 && (blockHeight-w.height) > txmp.config.TTLNumBlocks {
 			txmp.removeTxByElement(cur)
+			txmp.cache.Remove(w.tx)
 			txmp.metrics.EvictedTxs.Add(1)
 		} else if txmp.config.TTLDuration > 0 && now.Sub(w.timestamp) > txmp.config.TTLDuration {
 			txmp.removeTxByElement(cur)
+			txmp.cache.Remove(w.tx)
 			txmp.metrics.EvictedTxs.Add(1)
 		}
 		cur = next

internal/mempool/v1/mempool_test.go

@@ -209,7 +209,7 @@ func TestTxMempool_Size(t *testing.T) {
 }
 
 func TestTxMempool_Eviction(t *testing.T) {
-	txmp := setup(t, 0)
+	txmp := setup(t, 1000)
 	txmp.config.Size = 5
 	txmp.config.MaxTxsBytes = 60
 	txExists := func(spec string) bool {
@@ -238,6 +238,7 @@ func TestTxMempool_Eviction(t *testing.T) {
 	mustCheckTx(t, txmp, "key1=0000=25")
 	require.True(t, txExists("key1=0000=25"))
 	require.False(t, txExists(bigTx))
+	require.False(t, txmp.cache.Has([]byte(bigTx)))
 	require.Equal(t, int64(len("key1=0000=25")), txmp.SizeBytes())
 
 	// Now fill up the rest of the slots with other transactions.
@@ -521,10 +522,10 @@ func TestTxMempool_ConcurrentTxs(t *testing.T) {
 }
 
 func TestTxMempool_ExpiredTxs_Timestamp(t *testing.T) {
-	txmp := setup(t, 50)
+	txmp := setup(t, 5000)
 	txmp.config.TTLDuration = 5 * time.Millisecond
 
-	added1 := checkTxs(t, txmp, 25, 0)
+	added1 := checkTxs(t, txmp, 10, 0)
 	require.Equal(t, len(added1), txmp.Size())
 
 	// Wait a while, then add some more transactions that should not be expired
@@ -540,7 +541,7 @@ func TestTxMempool_ExpiredTxs_Timestamp(t *testing.T) {
 	// The exact intervals are not important except that the delta should be
 	// large relative to the cost of CheckTx (ms vs. ns is fine here).
 	time.Sleep(3 * time.Millisecond)
-	added2 := checkTxs(t, txmp, 25, 1)
+	added2 := checkTxs(t, txmp, 10, 1)
 
 	// Wait a while longer, so that the first batch will expire.
 	time.Sleep(3 * time.Millisecond)
@@ -555,6 +556,9 @@ func TestTxMempool_ExpiredTxs_Timestamp(t *testing.T) {
 		if _, ok := txmp.txByKey[tx.tx.Key()]; ok {
 			t.Errorf("Transaction %X should have been purged for TTL", tx.tx.Key())
 		}
+		if txmp.cache.Has(tx.tx) {
+			t.Errorf("Transaction %X should have been removed from the cache", tx.tx.Key())
+		}
 	}
 
 	// All the transactions added later should still be around.

node/node.go

@@ -250,7 +250,7 @@ func makeNode(cfg *config.Config,
 
 	// Determine whether we should do block sync. This must happen after the handshake, since the
 	// app may modify the validator set, specifying ourself as the only validator.
-	blockSync := cfg.BlockSync.Enable && !onlyValidatorIsUs(state, pubKey)
+	blockSync := !onlyValidatorIsUs(state, pubKey)
 
 	logNodeStartupInfo(state, pubKey, logger, consensusLogger, cfg.Mode)
 
@@ -777,30 +777,25 @@ func (n *nodeImpl) OnStart() error {
 
 			n.consensusReactor.SetStateSyncingMetrics(0)
 
-			d := types.EventDataStateSyncStatus{Complete: true, Height: state.LastBlockHeight}
-			if err := n.eventBus.PublishEventStateSyncStatus(d); err != nil {
+			if err := n.eventBus.PublishEventStateSyncStatus(
+				types.EventDataStateSyncStatus{Complete: true, Height: state.LastBlockHeight},
+			); err != nil {
 				n.eventBus.Logger.Error("failed to emit the statesync start event", "err", err)
 			}
 
-			// TODO: Some form of orchestrator is needed here between the state
-			// advancing reactors to be able to control which one of the three
-			// is running
-			if n.config.BlockSync.Enable {
-				// FIXME Very ugly to have these metrics bleed through here.
-				n.consensusReactor.SetBlockSyncingMetrics(1)
-				if err := bcR.SwitchToBlockSync(state); err != nil {
-					n.Logger.Error("failed to switch to block sync", "err", err)
-					return
-				}
-
-				d := types.EventDataBlockSyncStatus{Complete: false, Height: state.LastBlockHeight}
-				if err := n.eventBus.PublishEventBlockSyncStatus(d); err != nil {
-					n.eventBus.Logger.Error("failed to emit the block sync starting event", "err", err)
-				}
-
-			} else {
-				n.consensusReactor.SwitchToConsensus(state, true)
+			// FIXME Very ugly to have these metrics bleed through here.
+			n.consensusReactor.SetBlockSyncingMetrics(1)
+			if err := bcR.SwitchToBlockSync(state); err != nil {
+				n.Logger.Error("failed to switch to block sync", "err", err)
+				return
 			}
+
+			if err := n.eventBus.PublishEventBlockSyncStatus(
+				types.EventDataBlockSyncStatus{Complete: false, Height: state.LastBlockHeight},
+			); err != nil {
+				n.eventBus.Logger.Error("failed to emit the block sync starting event", "err", err)
+			}
+
 		}()
 	}
 

node/setup.go

@@ -338,6 +338,10 @@ func createBlockchainReactor(
 	metrics *consensus.Metrics,
 ) (*p2p.ReactorShim, service.Service, error) {
 
+	if !cfg.BlockSync.Enable {
+		logger.Error("blocksync.enable = false, but Tendermint no longer allows blocksync to be disabled. This setting is now ignored and will be removed in the next version.")
+	}
+
 	logger = logger.With("module", "blockchain")
 
 	switch cfg.BlockSync.Version {
@@ -441,12 +445,21 @@ func createConsensusReactor(
 }
 
 func createTransport(logger log.Logger, cfg *config.Config) *p2p.MConnTransport {
+	var maxAccepted uint32
+	switch {
+	case cfg.P2P.MaxConnections > 0 && !cfg.P2P.UseLegacy:
+		maxAccepted = uint32(cfg.P2P.MaxConnections) +
+			uint32(len(tmstrings.SplitAndTrimEmpty(cfg.P2P.UnconditionalPeerIDs, ",", " ")))
+
+	case cfg.P2P.MaxNumInboundPeers > 0:
+		maxAccepted = uint32(cfg.P2P.MaxNumInboundPeers) +
+			uint32(len(tmstrings.SplitAndTrimEmpty(cfg.P2P.UnconditionalPeerIDs, ",", " ")))
+	}
+
 	return p2p.NewMConnTransport(
 		logger, p2p.MConnConfig(cfg.P2P), []*p2p.ChannelDescriptor{},
 		p2p.MConnTransportOptions{
-			MaxAcceptedConnections: uint32(cfg.P2P.MaxNumInboundPeers +
-				len(tmstrings.SplitAndTrimEmpty(cfg.P2P.UnconditionalPeerIDs, ",", " ")),
-			),
+			MaxAcceptedConnections: maxAccepted,
 		},
 	)
 }

test/e2e/generator/generate.go

@@ -47,8 +47,6 @@ var (
 		"tcp":  20,
 		"unix": 10,
 	}
-	// FIXME: v2 disabled due to flake
-	nodeBlockSyncs = uniformChoice{"v0"} // "v2"
 	nodeMempools   = uniformChoice{"v0", "v1"}
 	nodeStateSyncs = weightedChoice{
 		e2e.StateSyncDisabled: 10,
@@ -397,7 +395,7 @@ func generateNode(
 		StartAt:          startAt,
 		Database:         nodeDatabases.Choose(r),
 		PrivvalProtocol:  nodePrivvalProtocols.Choose(r),
-		BlockSync:        nodeBlockSyncs.Choose(r).(string),
+		BlockSync:        "v0",
 		Mempool:          nodeMempools.Choose(r).(string),
 		StateSync:        e2e.StateSyncDisabled,
 		PersistInterval:  ptrUint64(uint64(nodePersistIntervals.Choose(r).(int))),

test/e2e/networks/ci.toml

@@ -43,7 +43,6 @@ persist_interval = 0
 perturb = ["restart"]
 privval_protocol = "tcp"
 seeds = ["seed01"]
-block_sync = "v0"
 
 [node.validator03]
 database = "badgerdb"
@@ -52,7 +51,6 @@ abci_protocol = "grpc"
 persist_interval = 3
 perturb = ["kill"]
 privval_protocol = "grpc"
-block_sync = "v0"
 retain_blocks = 10
 
 [node.validator04]
@@ -61,11 +59,9 @@ snapshot_interval = 5
 database = "rocksdb"
 persistent_peers = ["validator01"]
 perturb = ["pause"]
-block_sync = "v0"
 
 [node.validator05]
 database = "cleveldb"
-block_sync = "v0"
 state_sync = "p2p"
 seeds = ["seed01"]
 start_at = 1005 # Becomes part of the validator set at 1010
@@ -76,7 +72,6 @@ privval_protocol = "tcp"
 [node.full01]
 mode = "full"
 start_at = 1010
-block_sync = "v0"
 persistent_peers = ["validator01", "validator02", "validator03", "validator04"]
 perturb = ["restart"]
 retain_blocks = 10

test/e2e/pkg/testnet.go

@@ -178,7 +178,7 @@ func LoadTestnet(file string) (*Testnet, error) {
 			ABCIProtocol:     Protocol(testnet.ABCIProtocol),
 			PrivvalProtocol:  ProtocolFile,
 			StartAt:          nodeManifest.StartAt,
-			BlockSync:        nodeManifest.BlockSync,
+			BlockSync:        "v0",
 			Mempool:          nodeManifest.Mempool,
 			StateSync:        nodeManifest.StateSync,
 			PersistInterval:  1,

version/version.go

@@ -10,7 +10,7 @@ const (
 
 	// TMVersionDefault is the used as the fallback version of Tendermint Core
 	// when not using git describe. It is formatted with semantic versioning.
-	TMVersionDefault = "0.35.7"
+	TMVersionDefault = "0.35.8"
 
 	// ABCISemVer is the semantic version of the ABCI library
 	ABCISemVer = "0.17.0"