Merge branch 'master' into fix-issue-4941

.github/workflows/check.yaml

@@ -9,7 +9,7 @@ jobs:
     steps:
       - uses: actions/setup-go@v2
         with:
-          go-version: 1.17
+          go-version: 1.18
       - name: Checkout code
         uses: actions/checkout@v2
       - name: Restore cache
@@ -18,7 +18,6 @@ jobs:
           path: |
             ~/go/pkg/mod
             ~/.cache/go-build
-            **/.tools
             **/.dashboard_download_cache
           key: ${{ runner.os }}-golang-${{ hashFiles('**/go.sum') }}
           restore-keys: |

Makefile

@@ -137,7 +137,7 @@ SHELL := env PATH='$(PATH)' GOBIN='$(GO_TOOLS_BIN_PATH)' $(shell which bash)
 
 install-tools:
 	@mkdir -p $(GO_TOOLS_BIN_PATH)
-	@which golangci-lint >/dev/null 2>&1 || curl -sSfL https://raw.githubusercontent.com/golangci/golangci-lint/master/install.sh | sh -s -- -b $(GO_TOOLS_BIN_PATH) v1.43.0
+	@which golangci-lint >/dev/null 2>&1 || curl -sSfL https://raw.githubusercontent.com/golangci/golangci-lint/master/install.sh | sh -s -- -b $(GO_TOOLS_BIN_PATH) v1.46.0
 	@grep '_' tools.go | sed 's/"//g' | awk '{print $$2}' | xargs go install
 
 .PHONY: install-tools

go.mod

@@ -50,6 +50,7 @@ require (
 	go.etcd.io/etcd v0.5.0-alpha.5.0.20191023171146-3cf2f69b5738
 	go.uber.org/goleak v1.1.12
 	go.uber.org/zap v1.19.1
+	golang.org/x/text v0.3.3
 	golang.org/x/time v0.0.0-20220224211638-0e9765cccd65
 	golang.org/x/tools v0.1.5
 	google.golang.org/grpc v1.26.0

metrics/grafana/pd.json

@@ -3277,6 +3277,95 @@
             "align": false,
             "alignLevel": null
           }
+        },{
+          "aliasColors": {},
+          "bars": false,
+          "dashLength": 10,
+          "dashes": false,
+          "datasource": "${DS_TEST-CLUSTER}",
+          "description": "99% operator region size",
+          "fill": 0,
+          "gridPos": {
+            "h": 8,
+            "w": 12,
+            "x": 12,
+            "y": 42
+          },
+          "id": 1455,
+          "legend": {
+            "alignAsTable": true,
+            "avg": true,
+            "current": true,
+            "hideEmpty": true,
+            "hideZero": true,
+            "max": true,
+            "min": false,
+            "rightSide": true,
+            "show": true,
+            "total": false,
+            "values": true
+          },
+          "lines": true,
+          "linewidth": 2,
+          "links": [],
+          "nullPointMode": "null",
+          "percentage": false,
+          "pointradius": 2,
+          "points": false,
+          "renderer": "flot",
+          "seriesOverrides": [],
+          "spaceLength": 10,
+          "stack": false,
+          "steppedLine": false,
+          "targets": [
+            {
+              "expr": "histogram_quantile(0.99, sum(rate(pd_schedule_operator_region_size_bucket{k8s_cluster=\"$k8s_cluster\", tidb_cluster=\"$tidb_cluster\"}[5m])) by (type,le))",
+              "format": "time_series",
+              "intervalFactor": 2,
+              "legendFormat": "{{type,le}}",
+              "refId": "A"
+            }
+          ],
+          "thresholds": [],
+          "timeFrom": null,
+          "timeRegions": [],
+          "timeShift": null,
+          "title": "99% operator region size",
+          "tooltip": {
+            "shared": true,
+            "sort": 0,
+            "value_type": "individual"
+          },
+          "type": "graph",
+          "xaxis": {
+            "buckets": null,
+            "mode": "time",
+            "name": null,
+            "show": true,
+            "values": []
+          },
+          "yaxes": [
+            {
+              "format": "decmbytes",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            },
+            {
+              "format": "short",
+              "label": null,
+              "logBase": 1,
+              "max": null,
+              "min": null,
+              "show": true
+            }
+          ],
+          "yaxis": {
+            "align": false,
+            "alignLevel": null
+          }
         }
       ],
       "repeat": null,

pkg/autoscaling/prometheus.go

@@ -97,7 +97,7 @@ func (prom *PrometheusQuerier) queryMetricsFromPrometheus(query string, timestam
 		return nil, errs.ErrPrometheusQuery.Wrap(err).FastGenWithCause()
 	}
 
-	if warnings != nil && len(warnings) > 0 {
+	if len(warnings) > 0 {
 		log.Warn("prometheus query returns with warnings", zap.Strings("warnings", warnings))
 	}
 

server/grpc_service.go

@@ -34,6 +34,7 @@ import (
 	"github.com/tikv/pd/pkg/tsoutil"
 	"github.com/tikv/pd/server/cluster"
 	"github.com/tikv/pd/server/core"
+	"github.com/tikv/pd/server/schedule/operator"
 	"github.com/tikv/pd/server/storage/endpoint"
 	"github.com/tikv/pd/server/storage/kv"
 	"github.com/tikv/pd/server/tso"
@@ -1646,6 +1647,7 @@ func scatterRegions(cluster *cluster.RaftCluster, regionsID []uint64, group stri
 		return 0, err
 	}
 	for _, op := range ops {
+		op.AttachKind(operator.OpAdmin)
 		if ok := cluster.GetOperatorController().AddOperator(op); !ok {
 			failures[op.RegionID()] = fmt.Errorf("region %v failed to add operator", op.RegionID())
 		}

server/schedule/metrics.go

@@ -34,6 +34,15 @@ var (
 			Buckets:   []float64{0.5, 1, 2, 4, 8, 16, 20, 40, 60, 90, 120, 180, 240, 300, 480, 600, 720, 900, 1200, 1800, 3600},
 		}, []string{"type"})
 
+	operatorSizeHist = prometheus.NewHistogramVec(
+		prometheus.HistogramOpts{
+			Namespace: "pd",
+			Subsystem: "schedule",
+			Name:      "operator_region_size",
+			Help:      "Bucketed histogram of the operator region size.",
+			Buckets:   prometheus.ExponentialBuckets(1, 2, 20), // 1MB~1TB
+		}, []string{"type"})
+
 	operatorWaitCounter = prometheus.NewCounterVec(
 		prometheus.CounterOpts{
 			Namespace: "pd",
@@ -84,4 +93,5 @@ func init() {
 	prometheus.MustRegister(operatorWaitCounter)
 	prometheus.MustRegister(scatterCounter)
 	prometheus.MustRegister(scatterDistributionCounter)
+	prometheus.MustRegister(operatorSizeHist)
 }

server/schedule/operator_controller.go

@@ -470,6 +470,7 @@ func (oc *OperatorController) addOperatorLocked(op *operator.Operator) bool {
 	}
 	oc.operators[regionID] = op
 	operatorCounter.WithLabelValues(op.Desc(), "start").Inc()
+	operatorSizeHist.WithLabelValues(op.Desc()).Observe(float64(op.ApproximateSize))
 	operatorWaitDuration.WithLabelValues(op.Desc()).Observe(op.ElapsedTime().Seconds())
 	opInfluence := NewTotalOpInfluence([]*operator.Operator{op}, oc.cluster)
 	for storeID := range opInfluence.StoresInfluence {

tests/client/client_test.go

@@ -1292,7 +1292,8 @@ func (s *testConfigTTLSuite) TestConfigTTLAfterTransferLeader(c *C) {
 	addr := fmt.Sprintf("%s/pd/api/v1/config?ttlSecond=5", leader.GetAddr())
 	postData, err := json.Marshal(ttlConfig)
 	c.Assert(err, IsNil)
-	_, err = leader.GetHTTPClient().Post(addr, "application/json", bytes.NewBuffer(postData))
+	resp, err := leader.GetHTTPClient().Post(addr, "application/json", bytes.NewBuffer(postData))
+	resp.Body.Close()
 	c.Assert(err, IsNil)
 	time.Sleep(2 * time.Second)
 	_ = leader.Destroy()

tools/pd-ctl/pdctl/command/store_command.go

@@ -25,6 +25,8 @@ import (
 	"github.com/pingcap/kvproto/pkg/metapb"
 	"github.com/spf13/cobra"
 	"github.com/tikv/pd/server/api"
+	"golang.org/x/text/cases"
+	"golang.org/x/text/language"
 )
 
 var (
@@ -603,7 +605,8 @@ func storeCheckCommandFunc(cmd *cobra.Command, args []string) {
 		return
 	}
 
-	state := strings.Title(strings.ToLower(args[0]))
+	caser := cases.Title(language.Und)
+	state := caser.String(strings.ToLower(args[0]))
 	stateValue, ok := metapb.StoreState_value[state]
 	if !ok {
 		cmd.Println("Unknown state: " + state)