vttablet: handle applier metadata init failures in relay-log recovery

`handleRelayLogError` currently retries replication restart for known
recoverable metadata-init failures (relay log info and master info).
MySQL can also return:

```
Replica failed to initialize applier metadata structure from the repository
```

This treats this error as the same recoverable class by triggering
`RestartReplication` (STOP REPLICA, RESET REPLICA, START REPLICA).

Signed-off-by: Mohamed Hamza <mhamza@fastmail.com>

Author: mhamza15

go/vt/vttablet/tabletmanager/rpc_replication.go

@@ -20,6 +20,7 @@ import (
 	"context"
 	"fmt"
 	"runtime"
+	"slices"
 	"strings"
 	"time"
 
@@ -1237,6 +1238,35 @@ func (tm *TabletManager) fixSemiSyncAndReplication(ctx context.Context, tabletTy
 	return nil
 }
 
+// Known MySQL replication metadata initialization failures that can be repaired
+// by restarting replication.
+const (
+	relayLogInfoInitializationError    = "Replica failed to initialize relay log info structure from the repository"
+	masterInfoInitializationError      = "Could not initialize master info structure"
+	applierMetadataInitializationError = "Replica failed to initialize applier metadata structure from the repository"
+)
+
+// recoverableReplicationInitializationErrors enumerates the error substrings we
+// treat as recoverable through RestartReplication.
+var recoverableReplicationInitializationErrors = []string{
+	relayLogInfoInitializationError,
+	masterInfoInitializationError,
+	applierMetadataInitializationError,
+}
+
+// isRecoverableReplicationInitializationError returns true if err contains one
+// of the known recoverable metadata initialization failures.
+func isRecoverableReplicationInitializationError(err error) bool {
+	if err == nil {
+		return false
+	}
+
+	errMessage := err.Error()
+	return slices.ContainsFunc(recoverableReplicationInitializationErrors, func(s string) bool {
+		return strings.Contains(errMessage, s)
+	})
+}
+
 // handleRelayLogError resets replication of the instance.
 // This is required because sometimes MySQL gets stuck due to improper initialization of
 // master info structure or related failures and throws errors like
@@ -1247,8 +1277,7 @@ func (tm *TabletManager) handleRelayLogError(ctx context.Context, err error) err
 	// Replica failed to initialize relay log info structure from the repository (errno 1872) (sqlstate HY000) during query: START REPLICA
 	// see https://bugs.mysql.com/bug.php?id=83713 or https://github.com/vitessio/vitess/issues/5067
 	// The same fix also works for https://github.com/vitessio/vitess/issues/10955.
-	if strings.Contains(err.Error(), "Replica failed to initialize relay log info structure from the repository") ||
-		strings.Contains(err.Error(), "Could not initialize master info structure") {
+	if isRecoverableReplicationInitializationError(err) {
 		// Stop, reset and start replication again to resolve this error
 		if err := tm.MysqlDaemon.RestartReplication(ctx, tm.hookExtraEnv()); err != nil {
 			return err

go/vt/vttablet/tabletmanager/rpc_replication_test.go

@@ -18,6 +18,7 @@ package tabletmanager
 
 import (
 	"context"
+	"errors"
 	"sync/atomic"
 	"testing"
 	"time"
@@ -326,3 +327,65 @@ func TestUndoDemotePrimaryStateChange(t *testing.T) {
 	require.NoError(t, err)
 	require.False(t, isReadOnly)
 }
+
+func TestHandleRelayLogError(t *testing.T) {
+	testCases := []struct {
+		name          string
+		inputErr      error
+		shouldRestart bool
+	}{
+		{
+			name:          "relay log info error",
+			inputErr:      errors.New(relayLogInfoInitializationError),
+			shouldRestart: true,
+		},
+		{
+			name:          "master info error",
+			inputErr:      errors.New(masterInfoInitializationError),
+			shouldRestart: true,
+		},
+		{
+			name:          "applier metadata error",
+			inputErr:      errors.New(applierMetadataInitializationError),
+			shouldRestart: true,
+		},
+		{
+			name:          "unrelated error",
+			inputErr:      errors.New("unexpected replication failure"),
+			shouldRestart: false,
+		},
+	}
+
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			fakeMysqlDaemon := newTestMysqlDaemon(t, 1)
+			if tc.shouldRestart {
+				fakeMysqlDaemon.ExpectedExecuteSuperQueryList = []string{
+					"STOP REPLICA",
+					"RESET REPLICA",
+					"START REPLICA",
+				}
+			}
+
+			tablet := newTestTablet(t, 100, "ks", "0", nil)
+			tm := &TabletManager{
+				MysqlDaemon: fakeMysqlDaemon,
+				tabletAlias: tablet.Alias,
+				tmState: &tmState{
+					displayState: displayState{
+						tablet: tablet,
+					},
+				},
+			}
+
+			err := tm.handleRelayLogError(context.Background(), tc.inputErr)
+			if tc.shouldRestart {
+				require.NoError(t, err)
+			} else {
+				require.ErrorIs(t, err, tc.inputErr)
+			}
+
+			require.NoError(t, fakeMysqlDaemon.CheckSuperQueryList())
+		})
+	}
+}