ddl: fix ErrGCTooEarly when adding an index to the partition table (#14029)

ddl/failtest/fail_db_test.go

@@ -400,3 +400,24 @@ func (s *testFailDBSuite) TestRunDDLJobPanic(c *C) {
 	c.Assert(err, NotNil)
 	c.Assert(err.Error(), Equals, "[ddl:12]cancelled DDL job")
 }
+
+func (s *testFailDBSuite) TestPartitionAddIndexGC(c *C) {
+	tk := testkit.NewTestKit(c, s.store)
+	tk.MustExec("use test")
+	tk.MustExec(`create table partition_add_idx (
+	id int not null,
+	hired date not null
+	)
+	partition by range( year(hired) ) (
+	partition p1 values less than (1991),
+	partition p5 values less than (2008),
+	partition p7 values less than (2018)
+	);`)
+	tk.MustExec("insert into partition_add_idx values(1, '2010-01-01'), (2, '1990-01-01'), (3, '2001-01-01')")
+
+	c.Assert(failpoint.Enable("github.com/pingcap/tidb/ddl/mockUpdateCachedSafePoint", `return(true)`), IsNil)
+	defer func() {
+		c.Assert(failpoint.Disable("github.com/pingcap/tidb/ddl/mockUpdateCachedSafePoint"), IsNil)
+	}()
+	tk.MustExec("alter table partition_add_idx add index idx (id, hired)")
+}

ddl/index.go

@@ -36,6 +36,7 @@ import (
 	"github.com/pingcap/tidb/sessionctx"
 	"github.com/pingcap/tidb/sessionctx/variable"
 	"github.com/pingcap/tidb/store/tikv"
+	"github.com/pingcap/tidb/store/tikv/oracle"
 	"github.com/pingcap/tidb/table"
 	"github.com/pingcap/tidb/table/tables"
 	"github.com/pingcap/tidb/tablecodec"
@@ -1423,7 +1424,20 @@ func (w *worker) updateReorgInfo(t table.PartitionedTable, reorg *reorgInfo) (bo
 		return true, nil
 	}
 
-	start, end, err := getTableRange(reorg.d, t.GetPartition(pid), reorg.Job.SnapshotVer, reorg.Job.Priority)
+	failpoint.Inject("mockUpdateCachedSafePoint", func(val failpoint.Value) {
+		if val.(bool) {
+			// 18 is for the logical time.
+			ts := oracle.GetPhysical(time.Now()) << 18
+			s := reorg.d.store.(tikv.Storage)
+			s.UpdateSPCache(uint64(ts), time.Now())
+			time.Sleep(time.Millisecond * 3)
+		}
+	})
+	currentVer, err := getValidCurrentVersion(reorg.d.store)
+	if err != nil {
+		return false, errors.Trace(err)
+	}
+	start, end, err := getTableRange(reorg.d, t.GetPartition(pid), currentVer.Ver, reorg.Job.Priority)
 	if err != nil {
 		return false, errors.Trace(err)
 	}

ddl/reorg.go

@@ -340,9 +340,18 @@ func getTableRange(d *ddlCtx, tbl table.PhysicalTable, snapshotVer uint64, prior
 	return
 }
 
+func getValidCurrentVersion(store kv.Storage) (ver kv.Version, err error) {
+	ver, err = store.CurrentVersion()
+	if err != nil {
+		return ver, errors.Trace(err)
+	} else if ver.Ver <= 0 {
+		return ver, errInvalidStoreVer.GenWithStack("invalid storage current version %d", ver.Ver)
+	}
+	return ver, nil
+}
+
 func getReorgInfo(d *ddlCtx, t *meta.Meta, job *model.Job, tbl table.Table) (*reorgInfo, error) {
 	var (
-		err   error
 		start int64
 		end   int64
 		pid   int64
@@ -352,12 +361,9 @@ func getReorgInfo(d *ddlCtx, t *meta.Meta, job *model.Job, tbl table.Table) (*re
 	if job.SnapshotVer == 0 {
 		info.first = true
 		// get the current version for reorganization if we don't have
-		var ver kv.Version
-		ver, err = d.store.CurrentVersion()
+		ver, err := getValidCurrentVersion(d.store)
 		if err != nil {
 			return nil, errors.Trace(err)
-		} else if ver.Ver <= 0 {
-			return nil, errInvalidStoreVer.GenWithStack("invalid storage current version %d", ver.Ver)
 		}
 		tblInfo := tbl.Meta()
 		pid = tblInfo.ID
@@ -384,6 +390,7 @@ func getReorgInfo(d *ddlCtx, t *meta.Meta, job *model.Job, tbl table.Table) (*re
 		// Update info should after data persistent.
 		job.SnapshotVer = ver.Ver
 	} else {
+		var err error
 		start, end, pid, err = t.GetDDLReorgHandle(job)
 		if err != nil {
 			return nil, errors.Trace(err)
@@ -395,7 +402,7 @@ func getReorgInfo(d *ddlCtx, t *meta.Meta, job *model.Job, tbl table.Table) (*re
 	info.EndHandle = end
 	info.PhysicalTableID = pid
 
-	return &info, errors.Trace(err)
+	return &info, nil
 }
 
 func (r *reorgInfo) UpdateReorgMeta(txn kv.Transaction, startHandle, endHandle, physicalTableID int64) error {