server: add pagination to hot ranges API

This change extends hot ranges API to support pagination of responses. It is also possible to avoid pagination when page size is set to 0. Release note: None Release justification: bug fixes and low-risk updates to new functionality
cockroachdb · Mar 2, 2022 · af62e80 · af62e80
1 parent 35d2b78
commit af62e80
Show file tree

Hide file tree

Showing 5 changed files with 156 additions and 50 deletions.
diff --git a/docs/generated/http/full.md b/docs/generated/http/full.md
@@ -3130,6 +3130,8 @@ of ranges currently considered “hot” by the node(s).
 | Field | Type | Label | Description | Support status |
 | ----- | ---- | ----- | ----------- | -------------- |
 | node_id | [string](#cockroach.server.serverpb.HotRangesRequest-string) |  | NodeID indicates which node to query for a hot range report. It is possible to populate any node ID; if the node receiving the request is not the target node, it will forward the request to the target node.<br><br>If left empty, the request is forwarded to every node in the cluster. | [alpha](#support-status) |
+| page_size | [int32](#cockroach.server.serverpb.HotRangesRequest-int32) |  |  | [reserved](#support-status) |
+| page_token | [string](#cockroach.server.serverpb.HotRangesRequest-string) |  |  | [reserved](#support-status) |
 
 
 
@@ -3236,6 +3238,8 @@ of ranges currently considered “hot” by the node(s).
 | Field | Type | Label | Description | Support status |
 | ----- | ---- | ----- | ----------- | -------------- |
 | node_id | [string](#cockroach.server.serverpb.HotRangesRequest-string) |  | NodeID indicates which node to query for a hot range report. It is possible to populate any node ID; if the node receiving the request is not the target node, it will forward the request to the target node.<br><br>If left empty, the request is forwarded to every node in the cluster. | [alpha](#support-status) |
+| page_size | [int32](#cockroach.server.serverpb.HotRangesRequest-int32) |  |  | [reserved](#support-status) |
+| page_token | [string](#cockroach.server.serverpb.HotRangesRequest-string) |  |  | [reserved](#support-status) |
 
 
 
@@ -3253,7 +3257,9 @@ HotRangesResponseV2 is a response payload returned by `HotRangesV2` service.
 
 | Field | Type | Label | Description | Support status |
 | ----- | ---- | ----- | ----------- | -------------- |
-| ranges | [HotRangesResponseV2.HotRange](#cockroach.server.serverpb.HotRangesResponseV2-cockroach.server.serverpb.HotRangesResponseV2.HotRange) | repeated | ranges contain list of hot ranges info that has highest number of QPS. | [reserved](#support-status) |
+| ranges | [HotRangesResponseV2.HotRange](#cockroach.server.serverpb.HotRangesResponseV2-cockroach.server.serverpb.HotRangesResponseV2.HotRange) | repeated | Ranges contain list of hot ranges info that has highest number of QPS. | [reserved](#support-status) |
+| errors_by_node_id | [HotRangesResponseV2.ErrorsByNodeIdEntry](#cockroach.server.serverpb.HotRangesResponseV2-cockroach.server.serverpb.HotRangesResponseV2.ErrorsByNodeIdEntry) | repeated | errors contains any errors that occurred during fan-out calls to other nodes. | [reserved](#support-status) |
+| next_page_token | [string](#cockroach.server.serverpb.HotRangesResponseV2-string) |  | NextPageToken represents next pagination token to request next slice of data. | [reserved](#support-status) |
 
 
 
@@ -3281,6 +3287,20 @@ HotRange message describes a single hot range, ie its QPS, node ID it belongs to
 
 
 
+<a name="cockroach.server.serverpb.HotRangesResponseV2-cockroach.server.serverpb.HotRangesResponseV2.ErrorsByNodeIdEntry"></a>
+#### HotRangesResponseV2.ErrorsByNodeIdEntry
+
+
+
+| Field | Type | Label | Description | Support status |
+| ----- | ---- | ----- | ----------- | -------------- |
+| key | [int32](#cockroach.server.serverpb.HotRangesResponseV2-int32) |  |  |  |
+| value | [string](#cockroach.server.serverpb.HotRangesResponseV2-string) |  |  |  |
+
+
+
+
+
 
 ## Range
 

diff --git a/docs/generated/http/hotranges-request.md b/docs/generated/http/hotranges-request.md
@@ -12,5 +12,7 @@ Support status: [alpha](#support-status)
 | Field | Type | Label | Description | Support status |
 | ----- | ---- | ----- | ----------- | -------------- |
 | node_id | [string](#string) |  | NodeID indicates which node to query for a hot range report. It is possible to populate any node ID; if the node receiving the request is not the target node, it will forward the request to the target node.<br><br>If left empty, the request is forwarded to every node in the cluster. | [alpha](#support-status) |
+| page_size | [int32](#int32) |  |  | [reserved](#support-status) |
+| page_token | [string](#string) |  |  | [reserved](#support-status) |
 
 
diff --git a/pkg/server/serverpb/status.proto b/pkg/server/serverpb/status.proto
@@ -1110,6 +1110,8 @@ message HotRangesRequest {
   // in the cluster.
   // API: PUBLIC ALPHA
   string node_id = 1 [(gogoproto.customname) = "NodeID"];
+  int32 page_size = 2 [(gogoproto.nullable) = true];
+  string page_token = 3 [(gogoproto.nullable) = true];
 }
 
 // HotRangesResponse is the payload produced in response
@@ -1233,8 +1235,16 @@ message HotRangesResponseV2 {
     // schema_name provides the name of schema (if exists) for table in current range.
     string schema_name = 9;
   }
-  // ranges contain list of hot ranges info that has highest number of QPS.
+  // Ranges contain list of hot ranges info that has highest number of QPS.
   repeated HotRange ranges = 1;
+  // errors contains any errors that occurred during fan-out calls to other nodes.
+  map<int32, string> errors_by_node_id = 2 [
+    (gogoproto.castkey) = "github.com/cockroachdb/cockroach/pkg/roachpb.NodeID",
+    (gogoproto.customname) = "ErrorsByNodeID",
+    (gogoproto.nullable) = false
+  ];
+  // NextPageToken represents next pagination token to request next slice of data.
+  string next_page_token = 3 [(gogoproto.nullable) = true];
 }
 
 message RangeRequest {

diff --git a/pkg/server/status.go b/pkg/server/status.go
@@ -2121,14 +2121,23 @@ type hotRangeReportMeta struct {
 func (s *statusServer) HotRangesV2(
 	ctx context.Context, req *serverpb.HotRangesRequest,
 ) (*serverpb.HotRangesResponseV2, error) {
-	resp, err := s.HotRanges(ctx, req)
-	if err != nil {
+	if _, err := s.privilegeChecker.requireAdminUser(ctx); err != nil {
 		return nil, err
 	}
 
+	size := int(req.PageSize)
+	start := paginationState{}
+
+	if len(req.PageToken) > 0 {
+		if err := start.UnmarshalText([]byte(req.PageToken)); err != nil {
+			return nil, err
+		}
+	}
+
 	rangeReportMetas := make(map[uint32]hotRangeReportMeta)
 	var descrs []catalog.Descriptor
-	if err = s.sqlServer.distSQLServer.CollectionFactory.Txn(
+	var err error
+	if err := s.sqlServer.distSQLServer.CollectionFactory.Txn(
 		ctx, s.sqlServer.internalExecutor, s.db,
 		func(ctx context.Context, txn *kv.Txn, descriptors *descs.Collection) error {
 			all, err := descriptors.GetAllDescriptors(ctx, txn)
@@ -2162,51 +2171,99 @@ func (s *statusServer) HotRangesV2(
 		rangeReportMetas[id] = meta
 	}
 
-	var ranges []*serverpb.HotRangesResponseV2_HotRange
-	for nodeID, hr := range resp.HotRangesByNodeID {
-		for _, store := range hr.Stores {
-			for _, r := range store.HotRanges {
-				var (
-					dbName, tableName, indexName, schemaName string
-					replicaNodeIDs                           []roachpb.NodeID
-				)
-				_, tableID, err := s.sqlServer.execCfg.Codec.DecodeTablePrefix(r.Desc.StartKey.AsRawKey())
-				if err != nil {
-					log.Warningf(ctx, "cannot decode tableID for range descriptor: %s. %s", r.Desc.String(), err.Error())
-					continue
-				}
-				parent := rangeReportMetas[tableID].parentID
-				if parent != 0 {
-					tableName = rangeReportMetas[tableID].tableName
-					dbName = rangeReportMetas[parent].dbName
-				} else {
-					dbName = rangeReportMetas[tableID].dbName
-				}
-				schemaParent := rangeReportMetas[tableID].schemaParentID
-				schemaName = rangeReportMetas[schemaParent].schemaName
-				_, _, idxID, err := s.sqlServer.execCfg.Codec.DecodeIndexPrefix(r.Desc.StartKey.AsRawKey())
-				if err == nil {
-					indexName = rangeReportMetas[tableID].indexNames[idxID]
-				}
-				for _, repl := range r.Desc.Replicas().Descriptors() {
-					replicaNodeIDs = append(replicaNodeIDs, repl.NodeID)
+	response := &serverpb.HotRangesResponseV2{
+		ErrorsByNodeID: make(map[roachpb.NodeID]string),
+	}
+
+	var requestedNodes []roachpb.NodeID
+	if len(req.NodeID) > 0 {
+		requestedNodeID, _, err := s.parseNodeID(req.NodeID)
+		if err != nil {
+			return nil, err
+		}
+		requestedNodes = []roachpb.NodeID{requestedNodeID}
+	}
+
+	dialFn := func(ctx context.Context, nodeID roachpb.NodeID) (interface{}, error) {
+		client, err := s.dialNode(ctx, nodeID)
+		return client, err
+	}
+	remoteRequest := serverpb.HotRangesRequest{NodeID: "local"}
+	nodeFn := func(ctx context.Context, client interface{}, nodeID roachpb.NodeID) (interface{}, error) {
+		status := client.(serverpb.StatusClient)
+		resp, err := status.HotRanges(ctx, &remoteRequest)
+		if err != nil || resp == nil {
+			return nil, err
+		}
+		var ranges []*serverpb.HotRangesResponseV2_HotRange
+		for nodeID, hr := range resp.HotRangesByNodeID {
+			for _, store := range hr.Stores {
+				for _, r := range store.HotRanges {
+					var (
+						dbName, tableName, indexName, schemaName string
+						replicaNodeIDs                           []roachpb.NodeID
+					)
+					_, tableID, err := s.sqlServer.execCfg.Codec.DecodeTablePrefix(r.Desc.StartKey.AsRawKey())
+					if err != nil {
+						log.Warningf(ctx, "cannot decode tableID for range descriptor: %s. %s", r.Desc.String(), err.Error())
+						continue
+					}
+					parent := rangeReportMetas[tableID].parentID
+					if parent != 0 {
+						tableName = rangeReportMetas[tableID].tableName
+						dbName = rangeReportMetas[parent].dbName
+					} else {
+						dbName = rangeReportMetas[tableID].dbName
+					}
+					schemaParent := rangeReportMetas[tableID].schemaParentID
+					schemaName = rangeReportMetas[schemaParent].schemaName
+					_, _, idxID, err := s.sqlServer.execCfg.Codec.DecodeIndexPrefix(r.Desc.StartKey.AsRawKey())
+					if err == nil {
+						indexName = rangeReportMetas[tableID].indexNames[idxID]
+					}
+					for _, repl := range r.Desc.Replicas().Descriptors() {
+						replicaNodeIDs = append(replicaNodeIDs, repl.NodeID)
+					}
+					ranges = append(ranges, &serverpb.HotRangesResponseV2_HotRange{
+						RangeID:           r.Desc.RangeID,
+						NodeID:            nodeID,
+						QPS:               r.QueriesPerSecond,
+						TableName:         tableName,
+						SchemaName:        schemaName,
+						DatabaseName:      dbName,
+						IndexName:         indexName,
+						ReplicaNodeIds:    replicaNodeIDs,
+						LeaseholderNodeID: r.LeaseholderNodeID,
+					})
 				}
-				ranges = append(ranges, &serverpb.HotRangesResponseV2_HotRange{
-					RangeID:           r.Desc.RangeID,
-					NodeID:            nodeID,
-					QPS:               r.QueriesPerSecond,
-					TableName:         tableName,
-					SchemaName:        schemaName,
-					DatabaseName:      dbName,
-					IndexName:         indexName,
-					ReplicaNodeIds:    replicaNodeIDs,
-					LeaseholderNodeID: r.LeaseholderNodeID,
-				})
 			}
 		}
+		return ranges, nil
+	}
+	responseFn := func(nodeID roachpb.NodeID, resp interface{}) {
+		if resp == nil {
+			return
+		}
+		hotRanges := resp.([]*serverpb.HotRangesResponseV2_HotRange)
+		response.Ranges = append(response.Ranges, hotRanges...)
 	}
+	errorFn := func(nodeID roachpb.NodeID, err error) {
+		response.ErrorsByNodeID[nodeID] = err.Error()
+	}
+
+	next, err := s.paginatedIterateNodes(
+		ctx, "hotRanges", size, start, requestedNodes, dialFn,
+		nodeFn, responseFn, errorFn)
 
-	return &serverpb.HotRangesResponseV2{Ranges: ranges}, nil
+	if err != nil {
+		return nil, err
+	}
+	var nextBytes []byte
+	if nextBytes, err = next.MarshalText(); err != nil {
+		return nil, err
+	}
+	response.NextPageToken = string(nextBytes)
+	return response, nil
 }
 
 func (s *statusServer) localHotRanges(ctx context.Context) serverpb.HotRangesResponse_NodeResponse {

diff --git a/pkg/ui/workspaces/db-console/src/views/reports/containers/hotranges/index.tsx b/pkg/ui/workspaces/db-console/src/views/reports/containers/hotranges/index.tsx
@@ -24,16 +24,33 @@ const HotRanges = (props: HotRangesProps) => {
   const [hotRanges, setHotRanges] = useState<
     cockroach.server.serverpb.HotRangesResponseV2["ranges"]
   >([]);
-  const requestHotRanges = useCallback(() => {
+  const [pageToken, setPageToken] = useState<string>("");
+  const pageSize = 50;
+
+  const refreshHotRanges = useCallback(() => {
+    setHotRanges([]);
+    setPageToken("");
+  }, []);
+
+  useEffect(() => {
     const request = cockroach.server.serverpb.HotRangesRequest.create({
       node_id: nodeId,
+      page_size: pageSize,
+      page_token: pageToken,
     });
     getHotRanges(request).then(response => {
-      setHotRanges(response.ranges);
+      if (response.ranges.length == 0) {
+        return;
+      }
+      setPageToken(response.next_page_token);
+      setHotRanges([...hotRanges, ...response.ranges]);
       setTime(moment());
     });
-  }, [nodeId]);
-  useEffect(requestHotRanges, [requestHotRanges, nodeId]);
+    // Avoid dispatching request when `hotRanges` list is updated.
+    // This effect should be triggered only when pageToken is changed.
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, [pageToken]);
+
   useEffect(() => {
     setNodeId(nodeIdParam);
   }, [nodeIdParam]);
@@ -46,7 +63,7 @@ const HotRanges = (props: HotRangesProps) => {
     >
       <span>{`Node ID: ${nodeId ?? "All nodes"}`}</span>
       <span>{`Time: ${time.toISOString()}`}</span>
-      <Button onClick={requestHotRanges} intent={"secondary"}>
+      <Button onClick={refreshHotRanges} intent={"secondary"}>
         Refresh
       </Button>
       <pre className="state-json-box">{JSON.stringify(hotRanges, null, 2)}</pre>