scheduler: hot write scheduler randomly select the balance strategy (#…

…1034)
tikv · Apr 19, 2018 · 5bc920f · 5bc920f
1 parent 986c3ef
commit 5bc920f
Show file tree

Hide file tree

Showing 2 changed files with 35 additions and 19 deletions.
diff --git a/server/schedulers/balance_test.go b/server/schedulers/balance_test.go
@@ -904,16 +904,24 @@ func (s *testBalanceHotWriteRegionSchedulerSuite) TestBalance(c *C) {
 	tc.addLeaderRegionWithWriteInfo(3, 1, 512*1024*schedule.RegionHeartBeatReportInterval, 2, 4)
 	opt.HotRegionLowThreshold = 0
 
-	// Will transfer a hot region from store 1 to store 6, because the total count of peers
+	// Will transfer a hot region from store 1, because the total count of peers
 	// which is hot for store 1 is more larger than other stores.
 	op := hb.Schedule(tc, schedule.NewOpInfluence(nil, tc))
 	c.Assert(op, NotNil)
-	if op[0].RegionID() == 2 {
-		checkTransferPeerWithLeaderTransferFrom(c, op[0], schedule.OpHotRegion, 1)
-	} else {
-		checkTransferPeerWithLeaderTransfer(c, op[0], schedule.OpHotRegion, 1, 6)
+	switch op[0].Len() {
+	case 1:
+		// balance by leader selected
+		checkTransferLeaderFrom(c, op[0], schedule.OpHotRegion, 1)
+	case 3:
+		// balance by peer selected
+		if op[0].RegionID() == 2 {
+			// peer in store 1 of the region 2 can transfer to store 5 or store 6 because of the label
+			checkTransferPeerWithLeaderTransferFrom(c, op[0], schedule.OpHotRegion, 1)
+		} else {
+			// peer in store 1 of the region 1,2 can only transfer to store 6
+			checkTransferPeerWithLeaderTransfer(c, op[0], schedule.OpHotRegion, 1, 6)
+		}
 	}
-
 	// After transfer a hot region from store 1 to store 5
 	//| region_id | leader_sotre | follower_store | follower_store | written_bytes |
 	//|-----------|--------------|----------------|----------------|---------------|

diff --git a/server/schedulers/hot_region.go b/server/schedulers/hot_region.go
@@ -161,20 +161,28 @@ func (h *balanceHotRegionsScheduler) balanceHotReadRegions(cluster schedule.Clus
 	return nil
 }
 
-func (h *balanceHotRegionsScheduler) balanceHotWriteRegions(cluster schedule.Cluster) []*schedule.Operator {
-	// balance by peer
-	srcRegion, srcPeer, destPeer := h.balanceByPeer(cluster, h.stats.writeStatAsPeer)
-	if srcRegion != nil {
-		schedulerCounter.WithLabelValues(h.GetName(), "move_peer").Inc()
-		return []*schedule.Operator{schedule.CreateMovePeerOperator("moveHotWriteRegion", cluster, srcRegion, schedule.OpHotRegion, srcPeer.GetStoreId(), destPeer.GetStoreId(), destPeer.GetId())}
-	}
+// balanceHotRetryLimit is the limit to retry schedule for selected balance strategy.
+const balanceHotRetryLimit = 10
 
-	// balance by leader
-	srcRegion, newLeader := h.balanceByLeader(cluster, h.stats.writeStatAsLeader)
-	if srcRegion != nil {
-		schedulerCounter.WithLabelValues(h.GetName(), "move_leader").Inc()
-		step := schedule.TransferLeader{FromStore: srcRegion.Leader.GetStoreId(), ToStore: newLeader.GetStoreId()}
-		return []*schedule.Operator{schedule.NewOperator("transferHotWriteLeader", srcRegion.GetId(), schedule.OpHotRegion|schedule.OpLeader, step)}
+func (h *balanceHotRegionsScheduler) balanceHotWriteRegions(cluster schedule.Cluster) []*schedule.Operator {
+	for i := 0; i < balanceHotRetryLimit; i++ {
+		switch h.r.Int() % 2 {
+		case 0:
+			// balance by peer
+			srcRegion, srcPeer, destPeer := h.balanceByPeer(cluster, h.stats.writeStatAsPeer)
+			if srcRegion != nil {
+				schedulerCounter.WithLabelValues(h.GetName(), "move_peer").Inc()
+				return []*schedule.Operator{schedule.CreateMovePeerOperator("moveHotWriteRegion", cluster, srcRegion, schedule.OpHotRegion, srcPeer.GetStoreId(), destPeer.GetStoreId(), destPeer.GetId())}
+			}
+		case 1:
+			// balance by leader
+			srcRegion, newLeader := h.balanceByLeader(cluster, h.stats.writeStatAsLeader)
+			if srcRegion != nil {
+				schedulerCounter.WithLabelValues(h.GetName(), "move_leader").Inc()
+				step := schedule.TransferLeader{FromStore: srcRegion.Leader.GetStoreId(), ToStore: newLeader.GetStoreId()}
+				return []*schedule.Operator{schedule.NewOperator("transferHotWriteLeader", srcRegion.GetId(), schedule.OpHotRegion|schedule.OpLeader, step)}
+			}
+		}
 	}
 
 	schedulerCounter.WithLabelValues(h.GetName(), "skip").Inc()