Wal file size bug fix (hyperledger#4340)

* WAL "max entry size limit exceeded" bug reproduction

Signed-off-by: Emil Elizarov <[email protected]>

* Upgraded etcd Raft 3.5.1 -> 3.5.9

Signed-off-by: Emil Elizarov <[email protected]>

---------

Signed-off-by: Emil Elizarov <[email protected]>
Co-authored-by: Emil Elizarov <[email protected]>

go.mod

@@ -35,9 +35,9 @@ require (
 	github.com/sykesm/zap-logfmt v0.0.4
 	github.com/syndtr/goleveldb v1.0.1-0.20210305035536-64b5b1c73954
 	github.com/tedsuo/ifrit v0.0.0-20220120221754-dd274de71113
-	go.etcd.io/etcd/client/pkg/v3 v3.5.1
-	go.etcd.io/etcd/raft/v3 v3.5.1
-	go.etcd.io/etcd/server/v3 v3.5.1
+	go.etcd.io/etcd/client/pkg/v3 v3.5.9
+	go.etcd.io/etcd/raft/v3 v3.5.9
+	go.etcd.io/etcd/server/v3 v3.5.9
 	go.uber.org/zap v1.19.0
 	golang.org/x/crypto v0.1.0
 	golang.org/x/tools v0.1.12
@@ -98,7 +98,7 @@ require (
 	github.com/spf13/jwalterweatherman v1.1.0 // indirect
 	github.com/stretchr/objx v0.5.0 // indirect
 	github.com/subosito/gotenv v1.2.0 // indirect
-	go.etcd.io/etcd/pkg/v3 v3.5.1 // indirect
+	go.etcd.io/etcd/pkg/v3 v3.5.9 // indirect
 	go.opencensus.io v0.22.4 // indirect
 	go.uber.org/atomic v1.7.0 // indirect
 	go.uber.org/multierr v1.6.0 // indirect

orderer/consensus/etcdraft/chain_test.go

@@ -249,6 +249,32 @@ var _ = Describe("Chain", func() {
 		})
 
 		Context("when a node starts up", func() {
+			It("WAL 32MB file size should be within size limits", func() {
+				close(cutter.Block)
+				cutter.SetCutNext(true)
+				// Wait for the leader to be chosen
+				campaign(chain, observeC)
+
+				By("Create large TX")
+				large_size_env := &common.Envelope{
+					Payload: marshalOrPanic(&common.Payload{
+						Header: &common.Header{ChannelHeader: marshalOrPanic(&common.ChannelHeader{Type: int32(common.HeaderType_MESSAGE), ChannelId: channelID})},
+						Data:   make([]byte, 32_000_000),
+					}),
+				}
+
+				By("Send TX to chain")
+				Expect(chain.Order(large_size_env, 0)).To(Succeed())
+				By("Wait for TX to be written to WAL")
+				Eventually(support.WriteBlockCallCount, LongEventualTimeout).Should(Equal(1))
+				By("Stop the chain")
+				chain.Halt()
+
+				By("Create new chain")
+				_, err := etcdraft.NewChain(support, opts, configurator, nil, cryptoProvider, noOpBlockPuller, nil, observeC)
+				Expect(err).NotTo(HaveOccurred())
+			})
+
 			It("properly configures the communication layer", func() {
 				expectedNodeConfig := nodeConfigFromMetadata(consenterMetadata)
 				Eventually(configurator.ConfigureCallCount, LongEventualTimeout).Should(Equal(1))

orderer/consensus/etcdraft/storage_test.go

@@ -69,7 +69,6 @@ func assertFileCount(t *testing.T, wal, snap int) {
 func TestOpenWAL(t *testing.T) {
 	t.Run("Last WAL file is broken", func(t *testing.T) {
 		setup(t)
-		defer clean(t)
 
 		// create 10 new wal files
 		for i := 0; i < 10; i++ {
@@ -80,7 +79,7 @@ func TestOpenWAL(t *testing.T) {
 			)
 		}
 		assertFileCount(t, 1, 0)
-		lasti, _ := store.ram.LastIndex() // it never returns err
+		store.ram.LastIndex() // it never returns err
 
 		// close current storage
 		err = store.Close()
@@ -104,10 +103,7 @@ func TestOpenWAL(t *testing.T) {
 		// create new storage
 		ram = raft.NewMemoryStorage()
 		store, err = CreateStorage(logger, walDir, snapDir, ram)
-		require.NoError(t, err)
-		lastI, _ := store.ram.LastIndex()
-		require.True(t, lastI > 0)     // we are still able to read some entries
-		require.True(t, lasti > lastI) // but less than before because some are broken
+		require.ErrorContains(t, err, "failed to create or read WAL: failed to read WAL and cannot repair")
 	})
 }