Merge pull request 'updated createdeployinstance logics' (#292) from tzwang/pcm-coordinator:master into master

Former-commit-id: 3080046eb7
This commit is contained in:
tzwang 2024-08-29 17:24:02 +08:00
commit 9ec43307d7
6 changed files with 97 additions and 22 deletions

View File

@ -182,7 +182,7 @@ type (
TaskDesc string `form:"taskDesc"`
ModelName string `form:"modelName"`
ModelType string `form:"modelType"`
AdapterClusterMap map[string]string `form:"adapterClusterMap"`
AdapterClusterMap map[string][]string `form:"adapterClusterMap"`
}
CreateDeployTaskResp {

View File

@ -973,7 +973,7 @@ service pcm {
get /inference/getDeployTasksByType (GetDeployTasksByTypeReq) returns (GetDeployTasksByTypeResp)
@handler CreateDeployTask
get /inference/createDeployTask (CreateDeployTaskReq) returns (CreateDeployTaskResp)
post /inference/createDeployTask (CreateDeployTaskReq) returns (CreateDeployTaskResp)
@handler GetAdaptersByModel
get /inference/getAdaptersByModel (GetAdaptersByModelReq) returns (GetAdaptersByModelResp)

View File

@ -1234,7 +1234,7 @@ func RegisterHandlers(server *rest.Server, serverCtx *svc.ServiceContext) {
Handler: inference.GetDeployTasksByTypeHandler(serverCtx),
},
{
Method: http.MethodGet,
Method: http.MethodPost,
Path: "/inference/createDeployTask",
Handler: inference.CreateDeployTaskHandler(serverCtx),
},

View File

@ -2,6 +2,9 @@ package inference
import (
"context"
"errors"
"gitlink.org.cn/JointCloud/pcm-coordinator/internal/scheduler/schedulers/option"
"strconv"
"gitlink.org.cn/JointCloud/pcm-coordinator/internal/svc"
"gitlink.org.cn/JointCloud/pcm-coordinator/internal/types"
@ -24,7 +27,78 @@ func NewCreateDeployTaskLogic(ctx context.Context, svcCtx *svc.ServiceContext) *
}
func (l *CreateDeployTaskLogic) CreateDeployTask(req *types.CreateDeployTaskReq) (resp *types.CreateDeployTaskResp, err error) {
// todo: add your logic here and delete this line
resp = &types.CreateDeployTaskResp{}
if len(req.AdapterClusterMap) == 0 {
return nil, errors.New("adapters are empty")
}
opt := &option.InferOption{
TaskName: req.TaskName,
ModelType: req.ModelType,
ModelName: req.ModelName,
Cmd: "",
}
taskId, err := l.svcCtx.Scheduler.AiStorages.SaveInferDeployTask(req.TaskName, req.ModelName, req.ModelType, req.TaskDesc)
if err != nil {
return nil, err
}
for aid, v := range req.AdapterClusterMap {
for _, cid := range v {
err = l.createDeployInstance(taskId, aid, cid, opt)
if err != nil {
return nil, err
}
}
}
return
}
func (l *CreateDeployTaskLogic) createDeployInstance(taskId int64, adapterId string, clusterId string, opt *option.InferOption) error {
cmap, found := l.svcCtx.Scheduler.AiService.InferenceAdapterMap[adapterId]
if !found {
return errors.New("adapterId not exist: " + adapterId)
}
iCluster, found := cmap[clusterId]
if !found {
return errors.New("clusterId not exist: " + clusterId)
}
insId, err := iCluster.CreateInferDeployInstance(l.ctx, opt)
if err != nil {
return err
}
aid, err := strconv.ParseInt(adapterId, 10, 64)
if err != nil {
return err
}
cid, err := strconv.ParseInt(clusterId, 10, 64)
if err != nil {
return err
}
adapterName, err := l.svcCtx.Scheduler.AiStorages.GetAdapterNameById(adapterId)
if err != nil {
return err
}
clusterName, err := l.svcCtx.Scheduler.AiStorages.GetClusterNameById(clusterId)
if err != nil {
return err
}
ins, err := iCluster.GetInferDeployInstance(l.ctx, insId)
if err != nil {
return err
}
_, err = l.svcCtx.Scheduler.AiStorages.SaveInferDeployInstance(taskId, ins.InstanceId, ins.InstanceName, aid, adapterName, cid, clusterName, ins.ModelName, ins.ModelType, ins.InferCard)
if err != nil {
return err
}
return nil
}

View File

@ -373,23 +373,24 @@ func (s *AiStorage) AddNoticeInfo(adapterId string, adapterName string, clusterI
}
}
func (s *AiStorage) SaveInferDeployInstance(instanceId string, instanceName string, adapterId int64,
func (s *AiStorage) SaveInferDeployInstance(taskId int64, instanceId string, instanceName string, adapterId int64,
adapterName string, clusterId int64, clusterName string, modelName string, modelType string, inferCard string) (int64, error) {
startTime := time.Now().Format(time.RFC3339)
// 构建主任务结构体
insModel := models.AiInferDeployInstance{
InstanceId: instanceId,
InstanceName: instanceName,
AdapterId: adapterId,
AdapterName: adapterName,
ClusterId: clusterId,
ClusterName: clusterName,
ModelName: modelName,
ModelType: modelType,
InferCard: inferCard,
Status: constants.Saved,
CreateTime: startTime,
UpdateTime: startTime,
DeployInstanceTaskId: taskId,
InstanceId: instanceId,
InstanceName: instanceName,
AdapterId: adapterId,
AdapterName: adapterName,
ClusterId: clusterId,
ClusterName: clusterName,
ModelName: modelName,
ModelType: modelType,
InferCard: inferCard,
Status: constants.Stopped,
CreateTime: startTime,
UpdateTime: startTime,
}
// 保存任务数据到数据库
tx := s.DbEngin.Table("ai_infer_deploy_instance").Create(&insModel)

View File

@ -6075,11 +6075,11 @@ type GetDeployTasksByTypeResp struct {
}
type CreateDeployTaskReq struct {
TaskName string `form:"taskName"`
TaskDesc string `form:"taskDesc"`
ModelName string `form:"modelName"`
ModelType string `form:"modelType"`
AdapterClusterMap map[string]string `form:"adapterClusterMap"`
TaskName string `form:"taskName"`
TaskDesc string `form:"taskDesc"`
ModelName string `form:"modelName"`
ModelType string `form:"modelType"`
AdapterClusterMap map[string][]string `form:"adapterClusterMap"`
}
type CreateDeployTaskResp struct {