diff --git a/br/pkg/lightning/common/retry.go b/br/pkg/lightning/common/retry.go
index 19afcd2f90c46..e08fde0c02984 100644
--- a/br/pkg/lightning/common/retry.go
+++ b/br/pkg/lightning/common/retry.go
@@ -79,6 +79,8 @@ var retryableErrorIDs = map[errors.ErrorID]struct{}{
 	ErrKVReadIndexNotReady.ID():   {},
 	ErrKVIngestFailed.ID():        {},
 	ErrKVRaftProposalDropped.ID(): {},
+	// litBackendCtxMgr.Register may return the error.
+	ErrCreatePDClient.ID(): {},
 	// during checksum coprocessor will transform error into driver error in handleCopResponse using ToTiDBErr
 	// met ErrRegionUnavailable on free-tier import during checksum, others hasn't met yet
 	drivererr.ErrRegionUnavailable.ID(): {},
diff --git a/pkg/ddl/index.go b/pkg/ddl/index.go
index f8b5dca53d97f..aff05bb2ca49d 100644
--- a/pkg/ddl/index.go
+++ b/pkg/ddl/index.go
@@ -1081,9 +1081,7 @@ func runReorgJobAndHandleErr(w *worker, d *ddlCtx, t *meta.Meta, job *model.Job,
 			// TODO(tangenta): get duplicate column and match index.
 			err = convertToKeyExistsErr(err, allIndexInfos[0], tbl.Meta())
 		}
-		if !errorIsRetryable(err, job) ||
-			// TODO: Remove this check make it can be retry. Related test is TestModifyColumnReorgInfo.
-			job.ReorgMeta.IsDistReorg {
+		if !errorIsRetryable(err, job) {
 			logutil.BgLogger().Warn("run add index job failed, convert job to rollback", zap.String("category", "ddl"), zap.String("job", job.String()), zap.Error(err))
 			ver, err = convertAddIdxJob2RollbackJob(d, t, job, tbl.Meta(), allIndexInfos, err)
 			if err1 := rh.RemoveDDLReorgHandle(job, reorgInfo.elements); err1 != nil {
diff --git a/pkg/disttask/framework/scheduler/manager.go b/pkg/disttask/framework/scheduler/manager.go
index 95c9aa99c9bba..759e70612ab25 100644
--- a/pkg/disttask/framework/scheduler/manager.go
+++ b/pkg/disttask/framework/scheduler/manager.go
@@ -22,6 +22,7 @@ import (
 
 	"github.com/pingcap/errors"
 	"github.com/pingcap/failpoint"
+	"github.com/pingcap/tidb/br/pkg/lightning/common"
 	"github.com/pingcap/tidb/pkg/config"
 	"github.com/pingcap/tidb/pkg/disttask/framework/proto"
 	"github.com/pingcap/tidb/pkg/domain/infosync"
@@ -416,13 +417,18 @@ func (m *Manager) removeHandlingTask(id int64) {
 }
 
 func (m *Manager) logErr(err error) {
-	logutil.Logger(m.logCtx).Error("task manager error", zap.Error(err), zap.Stack("stack"))
+	logutil.Logger(m.logCtx).Error("task manager met error", zap.Error(err), zap.Stack("stack"))
 }
 
 func (m *Manager) logErrAndPersist(err error, taskID int64) {
 	m.logErr(err)
+	// TODO: use interface if each business to retry
+	if common.IsRetryableError(err) || isRetryableError(err) {
+		return
+	}
 	err1 := m.taskTable.UpdateErrorToSubtask(m.id, taskID, err)
 	if err1 != nil {
 		logutil.Logger(m.logCtx).Error("update to subtask failed", zap.Error(err1), zap.Stack("stack"))
 	}
+	logutil.Logger(m.logCtx).Error("update error to subtask", zap.Int64("task-id", taskID), zap.Error(err1), zap.Stack("stack"))
 }
diff --git a/pkg/disttask/framework/scheduler/scheduler.go b/pkg/disttask/framework/scheduler/scheduler.go
index 6253a2c6cab8b..357c39433a3d5 100644
--- a/pkg/disttask/framework/scheduler/scheduler.go
+++ b/pkg/disttask/framework/scheduler/scheduler.go
@@ -474,13 +474,13 @@ func (s *BaseScheduler) onError(err error) {
 		return
 	}
 	err = errors.Trace(err)
-	logutil.Logger(s.logCtx).Error("onError", zap.Error(err))
+	logutil.Logger(s.logCtx).Error("onError", zap.Error(err), zap.Stack("stack"))
 	s.mu.Lock()
 	defer s.mu.Unlock()
 
 	if s.mu.err == nil {
 		s.mu.err = err
-		logutil.Logger(s.logCtx).Error("scheduler error", zap.Error(err))
+		logutil.Logger(s.logCtx).Error("scheduler met first error", zap.Error(err))
 	}
 
 	if s.mu.runtimeCancel != nil {
@@ -620,5 +620,8 @@ func (s *BaseScheduler) updateErrorToSubtask(ctx context.Context, taskID int64,
 			return true, s.taskTable.UpdateErrorToSubtask(s.id, taskID, err)
 		},
 	)
+	if err1 == nil {
+		logger.Warn("update error to subtask success", zap.Error(err))
+	}
 	return err1
 }