From 1b8a2338abba2d5a55a36648ca698599329b69e3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?=E5=B1=B1=E5=B2=9A?=
 <36239017+YuJuncen@users.noreply.github.com>
Date: Fri, 30 Dec 2022 15:08:17 +0800
Subject: [PATCH] restore: set config value to default when failed to get
 config from tikv (#36051)

close pingcap/tidb#36053
---
 br/pkg/conn/conn.go        |  8 +++++---
 br/pkg/conn/conn_test.go   | 35 +++++++++++++++++++++++++++++++++--
 br/pkg/task/restore.go     |  5 +----
 br/pkg/task/restore_raw.go |  5 +----
 4 files changed, 40 insertions(+), 13 deletions(-)

diff --git a/br/pkg/conn/conn.go b/br/pkg/conn/conn.go
index 5adbe0a33ab1c..fff775bf1c1d7 100644
--- a/br/pkg/conn/conn.go
+++ b/br/pkg/conn/conn.go
@@ -281,7 +281,8 @@ func (mgr *Mgr) GetTS(ctx context.Context) (uint64, error) {
 }
 
 // GetMergeRegionSizeAndCount returns the tikv config `coprocessor.region-split-size` and `coprocessor.region-split-key`.
-func (mgr *Mgr) GetMergeRegionSizeAndCount(ctx context.Context, client *http.Client) (uint64, uint64, error) {
+// returns the default config when failed.
+func (mgr *Mgr) GetMergeRegionSizeAndCount(ctx context.Context, client *http.Client) (uint64, uint64) {
 	regionSplitSize := DefaultMergeRegionSizeBytes
 	regionSplitKeys := DefaultMergeRegionKeyCount
 	type coprocessor struct {
@@ -310,9 +311,10 @@ func (mgr *Mgr) GetMergeRegionSizeAndCount(ctx context.Context, client *http.Cli
 		return nil
 	})
 	if err != nil {
-		return 0, 0, errors.Trace(err)
+		log.Warn("meet error when getting config from TiKV; using default", logutil.ShortError(err))
+		return DefaultMergeRegionSizeBytes, DefaultMergeRegionKeyCount
 	}
-	return regionSplitSize, regionSplitKeys, nil
+	return regionSplitSize, regionSplitKeys
 }
 
 // GetConfigFromTiKV get configs from all alive tikv stores.
diff --git a/br/pkg/conn/conn_test.go b/br/pkg/conn/conn_test.go
index 01ce8bc08203e..fc822fac123d9 100644
--- a/br/pkg/conn/conn_test.go
+++ b/br/pkg/conn/conn_test.go
@@ -292,6 +292,38 @@ func TestGetMergeRegionSizeAndCount(t *testing.T) {
 			regionSplitSize: DefaultMergeRegionSizeBytes,
 			regionSplitKeys: DefaultMergeRegionKeyCount,
 		},
+		{
+			stores: []*metapb.Store{
+				{
+					Id:    1,
+					State: metapb.StoreState_Up,
+					Labels: []*metapb.StoreLabel{
+						{
+							Key:   "engine",
+							Value: "tiflash",
+						},
+					},
+				},
+				{
+					Id:    2,
+					State: metapb.StoreState_Up,
+					Labels: []*metapb.StoreLabel{
+						{
+							Key:   "engine",
+							Value: "tikv",
+						},
+					},
+				},
+			},
+			content: []string{
+				"",
+				// Assuming the TiKV has failed due to some reason.
+				"",
+			},
+			// no tikv detected in this case
+			regionSplitSize: DefaultMergeRegionSizeBytes,
+			regionSplitKeys: DefaultMergeRegionKeyCount,
+		},
 		{
 			stores: []*metapb.Store{
 				{
@@ -388,8 +420,7 @@ func TestGetMergeRegionSizeAndCount(t *testing.T) {
 		httpCli := mockServer.Client()
 		mgr := &Mgr{PdController: &pdutil.PdController{}}
 		mgr.PdController.SetPDClient(pdCli)
-		rs, rk, err := mgr.GetMergeRegionSizeAndCount(ctx, httpCli)
-		require.NoError(t, err)
+		rs, rk := mgr.GetMergeRegionSizeAndCount(ctx, httpCli)
 		require.Equal(t, ca.regionSplitSize, rs)
 		require.Equal(t, ca.regionSplitKeys, rk)
 		mockServer.Close()
diff --git a/br/pkg/task/restore.go b/br/pkg/task/restore.go
index 83c22a29e61db..0a1ba11cad84e 100644
--- a/br/pkg/task/restore.go
+++ b/br/pkg/task/restore.go
@@ -494,10 +494,7 @@ func RunRestore(c context.Context, g glue.Glue, cmdName string, cfg *RestoreConf
 		// according to https://github.com/pingcap/tidb/issues/34167.
 		// we should get the real config from tikv to adapt the dynamic region.
 		httpCli := httputil.NewClient(mgr.GetTLSConfig())
-		mergeRegionSize, mergeRegionCount, err = mgr.GetMergeRegionSizeAndCount(ctx, httpCli)
-		if err != nil {
-			return errors.Trace(err)
-		}
+		mergeRegionSize, mergeRegionCount = mgr.GetMergeRegionSizeAndCount(ctx, httpCli)
 	}
 
 	keepaliveCfg.PermitWithoutStream = true
diff --git a/br/pkg/task/restore_raw.go b/br/pkg/task/restore_raw.go
index 6c15cd9989512..7b80ac18b4d87 100644
--- a/br/pkg/task/restore_raw.go
+++ b/br/pkg/task/restore_raw.go
@@ -80,10 +80,7 @@ func RunRestoreRaw(c context.Context, g glue.Glue, cmdName string, cfg *RestoreR
 		// according to https://github.com/pingcap/tidb/issues/34167.
 		// we should get the real config from tikv to adapt the dynamic region.
 		httpCli := httputil.NewClient(mgr.GetTLSConfig())
-		mergeRegionSize, mergeRegionCount, err = mgr.GetMergeRegionSizeAndCount(ctx, httpCli)
-		if err != nil {
-			return errors.Trace(err)
-		}
+		mergeRegionSize, mergeRegionCount = mgr.GetMergeRegionSizeAndCount(ctx, httpCli)
 	}
 
 	keepaliveCfg := GetKeepalive(&cfg.Config)