monitoring

Former-commit-id: bc8021349d340169ad469dde4587591143700370

api/desc/core/pcm-core.api

@@ -31,6 +31,23 @@ type (
         centerType string `json:"centerType"`
     }
 )
+
+type remoteResp {
+    code int `json:"code"`
+    message string `json:"message"`
+    data interface{} `json:"data"`
+}
+
+type (
+    clustersLoadReq {
+        adapterId int64 `form:"adapterId"`
+        clusterName string `form:"clusterName"`
+    }
+    clustersLoadResp {
+        data interface{} `json:"data"`
+    }
+)
+
 type (
     syncClusterLoadReq {
         clusterLoadRecords []ClusterLoadRecord `json:"clusterLoadRecords"`

api/desc/pcm.api

@@ -11,18 +11,18 @@ import (
 	"schedule/pcm-schedule.api"
 )
 
-info (
+info(
-	title:   "pcm api service"
+	title: "pcm api service"
-	desc:    "type desc here"
+	desc: "type desc here"
-	author:  "type author here"
+	author: "type author here"
-	email:   "type email here"
+	email: "type email here"
 	version: "type version here"
 )
 
 //core端接口
-@server (
+@server(
 	prefix: pcm/v1
-	group:  core
+	group: core
 )
 service pcm {
 	@doc "查询P端服务列表"
@@ -112,12 +112,16 @@ service pcm {
 	@doc "alert rules"
 	@handler alertRulesHandler
 	get /core/alert/rules
+
+	@doc "cluster resource load"
+	@handler clustersLoadHandler
+	get /core/cluster/load (clustersLoadReq) returns (clustersLoadResp)
 }
 
 //hpc二级接口
-@server (
+@server(
 	prefix: pcm/v1
-	group:  hpc
+	group: hpc
 )
 service pcm {
 	@doc "提交超算任务"
@@ -138,9 +142,9 @@ service pcm {
 }
 
 //cloud二级接口
-@server (
+@server(
 	prefix: pcm/v1
-	group:  cloud
+	group: cloud
 )
 service pcm {
 	@doc "云算任务列表"
@@ -177,9 +181,9 @@ service pcm {
 }
 
 //智算二级接口
-@server (
+@server(
 	prefix: pcm/v1
-	group:  ai
+	group: ai
 )
 service pcm {
 	@doc "查询数据集列表"
@@ -300,13 +304,13 @@ service pcm {
 	@doc "创建虚拟化任务"
 	@handler createVisualizationJobHandler
 	post /ai/CreateVisualizationJob (CreateVisualizationJobReq) returns (CreateVisualizationJobResp)
-/******************Visualization Job Method start*************************/
+	/******************Visualization Job Method start*************************/
 }
 
 //screen接口
-@server (
+@server(
 	prefix: pcm/v1
-	group:  storage
+	group: storage
 )
 service pcm {
 	@doc "存储概览"
@@ -323,9 +327,9 @@ service pcm {
 }
 
 //镜像接口
-@server (
+@server(
 	prefix: pcm/v1
-	group:  image
+	group: image
 )
 service pcm {
 	@doc "镜像上传"
@@ -350,9 +354,9 @@ service pcm {
 }
 
 //openstack 接口
-@server (
+@server(
 	prefix: pcm/v1
-	group:  vm
+	group: vm
 )
 service pcm {
 	@doc "openstack计算中心概览"
@@ -741,9 +745,9 @@ service pcm {
 }
 
 //存算联动 接口
-@server (
+@server(
 	prefix: pcm/v1
-	group:  storelink
+	group: storelink
 )
 service pcm {
 	@handler UploadLinkImageHandler
@@ -772,9 +776,9 @@ service pcm {
 }
 
 // 接口
-@server (
+@server(
 	prefix: pcm/v1
-	group:  apps
+	group: apps
 )
 service pcm {
 	@doc "应用列表"
@@ -815,9 +819,9 @@ service pcm {
 }
 
 // 接口
-@server (
+@server(
 	prefix: pcm/v1
-	group:  adapters
+	group: adapters
 )
 service pcm {
 	@handler AdaptersListHandler
@@ -857,9 +861,9 @@ service pcm {
 	get /adapter/clusterSum (clusterSumReq) returns (clusterSumReqResp)
 }
 
-@server (
+@server(
 	prefix: pcm/v1
-	group:  schedule
+	group: schedule
 )
 service pcm {
 	@handler ScheduleGetAiResourceTypesHandler
@@ -878,9 +882,9 @@ service pcm {
 	post /schedule/submit (ScheduleReq) returns (ScheduleResp)
 }
 
-@server (
+@server(
 	prefix: pcm/v1
-	group:  dictionary
+	group: dictionary
 )
 service pcm {
 	@handler GetDict
@@ -915,5 +919,4 @@ service pcm {
 
 	@handler ListDictItemByCode
 	get /dictItem/code/:dictCode (DictCodeReq) returns (PageResult)
-}
+}
-

api/etc/pcm.yaml

@@ -11,11 +11,10 @@ Redis:
   Host: 10.206.0.12:6379
   Pass: redisPW123
 
-Cache:
+Monitoring:
-  - Host: 10.206.0.12:6379
+  PromUrl: http://47.92.39.128:30877
-    Pass: redisPW123
+  AlertUrl: 47.92.39.128:32243
 
-PromUrl: http://47.92.39.128:30877
 
 # k8s rpc
 K8sNativeConf:

api/internal/handler/core/clustersloadhandler.go

@@ -0,0 +1,25 @@
+package core
+
+import (
+	"gitlink.org.cn/JointCloud/pcm-coordinator/pkg/repository/result"
+	"net/http"
+
+	"github.com/zeromicro/go-zero/rest/httpx"
+	"gitlink.org.cn/JointCloud/pcm-coordinator/api/internal/logic/core"
+	"gitlink.org.cn/JointCloud/pcm-coordinator/api/internal/svc"
+	"gitlink.org.cn/JointCloud/pcm-coordinator/api/internal/types"
+)
+
+func ClustersLoadHandler(svcCtx *svc.ServiceContext) http.HandlerFunc {
+	return func(w http.ResponseWriter, r *http.Request) {
+		var req types.ClustersLoadReq
+		if err := httpx.Parse(r, &req); err != nil {
+			httpx.ErrorCtx(r.Context(), w, err)
+			return
+		}
+
+		l := core.NewClustersLoadLogic(r.Context(), svcCtx)
+		resp, err := l.ClustersLoad(&req)
+		result.HttpResult(r, w, resp, err)
+	}
+}

api/internal/handler/core/listdomainresourcehandler.go

@@ -1,32 +0,0 @@
-/*
-
- Copyright (c) [2023] [pcm]
- [pcm-coordinator] is licensed under Mulan PSL v2.
- You can use this software according to the terms and conditions of the Mulan PSL v2.
- You may obtain a copy of Mulan PSL v2 at:
-          http://license.coscl.org.cn/MulanPSL2
- THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
- EITHER EXPaRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
- MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
- See the Mulan PSL v2 for more details.
-
-*/
-
-package core
-
-import (
-	"net/http"
-
-	"gitlink.org.cn/JointCloud/pcm-coordinator/pkg/repository/result"
-
-	"gitlink.org.cn/JointCloud/pcm-coordinator/api/internal/logic/core"
-	"gitlink.org.cn/JointCloud/pcm-coordinator/api/internal/svc"
-)
-
-func ListDomainResourceHandler(svcCtx *svc.ServiceContext) http.HandlerFunc {
-	return func(w http.ResponseWriter, r *http.Request) {
-		l := core.NewListDomainResourceLogic(r.Context(), svcCtx)
-		resp, err := l.ListDomainResource()
-		result.HttpResult(r, w, resp, err)
-	}
-}

api/internal/logic/core/clustersloadlogic.go

@@ -0,0 +1,50 @@
+package core
+
+import (
+	"context"
+	"github.com/zeromicro/go-zero/core/logx"
+	"gitlink.org.cn/JointCloud/pcm-coordinator/api/internal/svc"
+	"gitlink.org.cn/JointCloud/pcm-coordinator/api/internal/types"
+	"gitlink.org.cn/JointCloud/pcm-coordinator/pkg/tracker"
+	"time"
+)
+
+type ClustersLoadLogic struct {
+	logx.Logger
+	ctx    context.Context
+	svcCtx *svc.ServiceContext
+}
+
+func NewClustersLoadLogic(ctx context.Context, svcCtx *svc.ServiceContext) *ClustersLoadLogic {
+	return &ClustersLoadLogic{
+		Logger: logx.WithContext(ctx),
+		ctx:    ctx,
+		svcCtx: svcCtx,
+	}
+}
+
+func (l *ClustersLoadLogic) ClustersLoad(req *types.ClustersLoadReq) (resp *types.ClustersLoadResp, err error) {
+	//resp = &types.ClustersLoadResp{}
+	//var server string
+	//metrics := []string{"cluster_cpu_total", "cluster_cpu_utilisation", "cluster_memory_avail", "cluster_memory_total", "cluster_disk_avail", "cluster_disk_total"}
+	//var remoteResp types.RemoteResp
+	//l.svcCtx.DbEngin.Raw("select server from t_adapter where id = ?", req.AdapterId).Scan(&server)
+	//adapterResp, err := l.svcCtx.HttpClient.R().
+	//	SetQueryParamsFromValues(map[string][]string{
+	//		"clusterName": []string{req.ClusterName},
+	//		"metrics":     metrics,
+	//	}).
+	//	SetResult(&remoteResp).
+	//	ForceContentType("application/json").
+	//	Get(server + "/api/v1/monitoring/cluster/load")
+	////Get("http://47.92.39.128:31031/pcm/v1/adapter/cluster/list")
+	//if err != nil || adapterResp.IsError() {
+	//	return nil, err
+	//}
+	//resp.Data = remoteResp.Data
+	resp = &types.ClustersLoadResp{}
+	metrics := []string{"cluster_cpu_utilisation", "cluster_cpu_avail", "cluster_cpu_total", "cluster_memory_total", "cluster_memory_avail", "cluster_memory_utilisation", "cluster_memory_utilisation", "cluster_memory_avail", "cluster_memory_total"}
+	result := l.svcCtx.PromClient.GetNamedMetrics(metrics, time.Now(), tracker.ClusterOption{})
+	println(len(result))
+	return resp, nil
+}

api/internal/svc/servicecontext.go

@@ -21,6 +21,7 @@ import (
 	"github.com/aws/aws-sdk-go/service/s3/s3manager"
 	"github.com/docker/docker/client"
 	"github.com/go-redis/redis/v8"
+	"github.com/go-resty/resty/v2"
 	alert "github.com/prometheus/alertmanager/api/v2/client"
 	"github.com/robfig/cron/v3"
 	"github.com/zeromicro/go-zero/core/logx"
@@ -64,6 +65,7 @@ type ServiceContext struct {
 	ParticipantRpc  participantservice.ParticipantService
 	PromClient      tracker.Prometheus
 	AlertClient     *alert.AlertmanagerAPI
+	HttpClient      *resty.Client
 }
 
 func NewServiceContext(c config.Config) *ServiceContext {
@@ -76,11 +78,12 @@ func NewServiceContext(c config.Config) *ServiceContext {
 		S3ForcePathStyle: aws.Bool(true),                                                                  //使用路径样式而非虚拟主机样式,区别请参考:https://docs.aws.amazon.com/AmazonS3/latest/dev/VirtualHosting.html
 	})
 	promClient, err := tracker.NewPrometheus(c.Monitoring.PromUrl)
+
 	if err != nil {
 		logx.Errorf("InitPrometheus err: %v", err)
 		panic("InitSnowflake err")
 	}
-
+	httpClient := resty.New()
 	alertClient := tracker.NewAlertClient(c.Monitoring.AlertUrl)
 	if err != nil {
 		logx.Errorf("InitPrometheus err: %v", err)
@@ -144,5 +147,6 @@ func NewServiceContext(c config.Config) *ServiceContext {
 		Uploader:        uploader,
 		PromClient:      promClient,
 		AlertClient:     alertClient,
+		HttpClient:      httpClient,
 	}
 }

pkg/tracker/queryoptions.go

@@ -101,6 +101,7 @@ func (a AdapterOption) Apply(o *QueryOptions) {
 }
 
 type ClusterOption struct {
+	AdapterId   int
 	ClusterName string
 }
 

pkg/tracker/tracker.go

@@ -138,6 +138,7 @@ func (p Prometheus) GetNamedMetricsByTime(metrics []string, start, end string, s
 				End:   endTimestamp,
 				Step:  step,
 			}
+			p.client.Rules(context.Background())
 			value, _, err := p.client.QueryRange(context.Background(), makeExpr(metric, *opts), timeRange)
 			if err != nil {
 				parsedResp.Error = err.Error()