br: skip some warning during restore system schemas

br/cmd/br/cmd.go

@@ -36,6 +36,7 @@ var (
 		"*.*",
 		fmt.Sprintf("!%s.*", utils.TemporaryDBName("*")),
 		"!mysql.*",
+		"mysql.bind_info",
 		"mysql.user",
 		"mysql.db",
 		"mysql.tables_priv",

br/cmd/br/restore.go

@@ -14,6 +14,7 @@ import (
 	"github.com/pingcap/tidb/br/pkg/trace"
 	"github.com/pingcap/tidb/br/pkg/utils"
 	"github.com/pingcap/tidb/br/pkg/version/build"
+	"github.com/pingcap/tidb/pkg/config"
 	"github.com/pingcap/tidb/pkg/session"
 	"github.com/pingcap/tidb/pkg/util/metricsutil"
 	"github.com/spf13/cobra"
@@ -38,6 +39,9 @@ func runRestoreCommand(command *cobra.Command, cmdName string) error {
 		}
 	}
 
+	// have to skip grant table, in order to NotifyUpdatePrivilege in binary mode
+	config.GetGlobalConfig().Security.SkipGrantTable = true
+
 	ctx := GetDefaultContext()
 	if cfg.EnableOpenTracing {
 		var store *appdash.MemoryStore

br/pkg/restore/BUILD.bazel

@@ -49,7 +49,6 @@ go_library(
         "//br/pkg/utils/storewatch",
         "//br/pkg/version",
         "//pkg/bindinfo",
-        "//pkg/config",
         "//pkg/ddl",
         "//pkg/ddl/util",
         "//pkg/domain",

br/pkg/restore/client.go

@@ -45,7 +45,6 @@ import (
 	"github.com/pingcap/tidb/br/pkg/utils"
 	"github.com/pingcap/tidb/br/pkg/utils/iter"
 	"github.com/pingcap/tidb/br/pkg/version"
-	"github.com/pingcap/tidb/pkg/config"
 	ddlutil "github.com/pingcap/tidb/pkg/ddl/util"
 	"github.com/pingcap/tidb/pkg/domain"
 	"github.com/pingcap/tidb/pkg/domain/infosync"
@@ -3537,11 +3536,6 @@ func (rc *Client) InitFullClusterRestore(explicitFilter bool) {
 	rc.fullClusterRestore = !explicitFilter && rc.IsFull()
 
 	log.Info("full cluster restore", zap.Bool("value", rc.fullClusterRestore))
-
-	if rc.fullClusterRestore {
-		// have to skip grant table, in order to NotifyUpdatePrivilege
-		config.GetGlobalConfig().Security.SkipGrantTable = true
-	}
 }
 
 func (rc *Client) IsFullClusterRestore() bool {

br/pkg/restore/db_test.go

@@ -415,3 +415,17 @@ func TestGetExistedUserDBs(t *testing.T) {
 	dbs = restore.GetExistedUserDBs(dom)
 	require.Equal(t, 2, len(dbs))
 }
+
+// NOTICE: Once there is a new system table, BR needs to ensure that it is correctly classified:
+//
+// - IF it is an unrecoverable table, please add the table name into `unRecoverableTable`.
+// - IF it is an system privilege table, please add the table name into `sysPrivilegeTableMap`.
+// - IF it is an statistics table, please add the table name into `statsTables`.
+//
+// The above variables are in the file br/pkg/restore/systable_restore.go
+func TestMonitorTheSystemTableIncremental(t *testing.T) {
+	s := createRestoreSchemaSuite(t)
+	tk := testkit.NewTestKit(t, s.mock.Storage)
+	ret := tk.MustQuery("SELECT COUNT(*) FROM INFORMATION_SCHEMA.TABLES WHERE TABLE_SCHEMA = 'mysql'")
+	ret.Equal([][]interface{}{{55}})
+}

br/pkg/restore/systable_restore.go

@@ -27,13 +27,16 @@ const (
 )
 
 var statsTables = map[string]struct{}{
-	"stats_buckets":    {},
-	"stats_extended":   {},
-	"stats_feedback":   {},
-	"stats_fm_sketch":  {},
-	"stats_histograms": {},
-	"stats_meta":       {},
-	"stats_top_n":      {},
+	"stats_buckets":      {},
+	"stats_extended":     {},
+	"stats_feedback":     {},
+	"stats_fm_sketch":    {},
+	"stats_histograms":   {},
+	"stats_history":      {},
+	"stats_meta":         {},
+	"stats_meta_history": {},
+	"stats_table_locked": {},
+	"stats_top_n":        {},
 }
 
 var unRecoverableTable = map[string]struct{}{
@@ -49,6 +52,9 @@ var unRecoverableTable = map[string]struct{}{
 
 	// schema_index_usage has table id need to be rewrite.
 	"schema_index_usage": {},
+
+	// replace into view is not supported now
+	"tidb_mdl_view": {},
 }
 
 // tables in this map is restored when fullClusterRestore=true
@@ -149,11 +155,16 @@ func (rc *Client) ClearSystemUsers(ctx context.Context, resetUsers []string) err
 
 // RestoreSystemSchemas restores the system schema(i.e. the `mysql` schema).
 // Detail see https://github.com/pingcap/br/issues/679#issuecomment-762592254.
-func (rc *Client) RestoreSystemSchemas(ctx context.Context, f filter.Filter) error {
+func (rc *Client) RestoreSystemSchemas(ctx context.Context, f filter.Filter) (rerr error) {
 	sysDB := mysql.SystemDB
 
 	temporaryDB := utils.TemporaryDBName(sysDB)
-	defer rc.cleanTemporaryDatabase(ctx, sysDB)
+	defer func() {
+		// Don't clean the temporary database for next restore with checkpoint.
+		if rerr == nil {
+			rc.cleanTemporaryDatabase(ctx, sysDB)
+		}
+	}()
 
 	if !f.MatchSchema(sysDB) || !rc.withSysTable {
 		log.Debug("system database filtered out", zap.String("database", sysDB))
@@ -223,15 +234,11 @@ func (rc *Client) afterSystemTablesReplaced(ctx context.Context, tables []string
 	var err error
 	for _, table := range tables {
 		if table == "user" {
-			if rc.fullClusterRestore {
-				log.Info("privilege system table restored, please reconnect to make it effective")
-				err = multierr.Append(err, rc.dom.NotifyUpdatePrivilege())
+			if serr := rc.dom.NotifyUpdatePrivilege(); serr != nil {
+				log.Warn("failed to flush privileges, please manually execute `FLUSH PRIVILEGES`")
+				err = multierr.Append(err, berrors.ErrUnknown.Wrap(serr).GenWithStack("failed to flush privileges"))
 			} else {
-				// to make it compatible with older version
-				// todo: should we allow restore system table in non-fresh cluster in later br version?
-				// if we don't, we can check it at first place.
-				err = multierr.Append(err, errors.Annotatef(berrors.ErrUnsupportedSystemTable,
-					"restored user info may not take effect, until you should execute `FLUSH PRIVILEGES` manually"))
+				log.Info("privilege system table restored, please reconnect to make it effective")
 			}
 		} else if table == "bind_info" {
 			if serr := rc.db.se.Execute(ctx, bindinfo.StmtRemoveDuplicatedPseudoBinding); serr != nil {
@@ -277,12 +284,11 @@ func (rc *Client) replaceTemporaryTableToSystable(ctx context.Context, ti *model
 	//  1.5 ) (Optional) The UPDATE statement sometimes costs, the whole system tables restore step can be place into the restore pipeline.
 	//  2   ) Deprecate the origin interface for backing up statistics.
 	if isStatsTable(tableName) {
-		return berrors.ErrUnsupportedSystemTable.GenWithStack("restoring stats via `mysql` schema isn't support yet: " +
-			"the table ID is out-of-date and may corrupt existing statistics")
+		return nil
 	}
 
 	if isUnrecoverableTable(tableName) {
-		return berrors.ErrUnsupportedSystemTable.GenWithStack("restoring unsupported `mysql` schema table")
+		return nil
 	}
 
 	// Currently, we don't support restore resource group metadata, so we need to

br/tests/br_full_cluster_restore/run.sh

@@ -162,5 +162,14 @@ check_contains "count(*): 0"
 run_sql "select count(*) from mysql.role_edges where to_user='cloud_admin'"
 check_contains "count(*): 0"
 
+echo "--> full cluster restore with --filter, need to flush privileges"
+restart_services
+run_br restore full --filter "*.*" --filter "!__TiDB_BR_Temporary*.*" --filter "!mysql.*" --filter "mysql.bind_info" --filter "mysql.user" --filter "mysql.db" --filter "mysql.tables_priv" --filter "mysql.columns_priv" --filter "mysql.global_priv" --filter "mysql.global_grants" --filter "mysql.default_roles" --filter "mysql.role_edges" --filter "!sys.*" --filter "!INFORMATION_SCHEMA.*" --filter "!PERFORMANCE_SCHEMA.*" --filter "!METRICS_SCHEMA.*" --filter "!INSPECTION_SCHEMA.*" --with-sys-table --log-file $br_log_file -s "local://$backup_dir"
+# BR executes `FLUSH PRIVILEGES` already
+run_sql_as user1 "123456" "select count(*) from db1.t1"
+check_contains "count(*): 2"
+run_sql_as user1 "123456" "select count(*) from db2.t1"
+check_contains "count(*): 2"
+
 echo "clean up kept backup"
 rm -rf $backup_dir $incr_backup_dir

br/tests/br_full_ddl/run.sh

@@ -204,8 +204,8 @@ then
 else
   echo "TEST: [$TEST_NAME] fail due to stats are not equal"
   grep ERROR $LOG
-  cat $BACKUP_STAT | head -n 1000
-  cat $RESOTRE_STAT | head -n 1000
+  cat $BACKUP_STAT | tail -n 1000
+  cat $RESOTRE_STAT | tail -n 1000
   exit 1
 fi