Support registry resource

doc/usage_cn.md

@@ -33,12 +33,21 @@ Usage: paddlecloud <flags> <subcommand> <subcommand args>
 
 Subcommands:
 	commands         list all command names
+	delete           Delete the specify resource.
+	file             Simple file operations.
 	get              Print resources
 	help             describe subcommands and their syntax
 	kill             Stop the job. -rm will remove the job from history.
 	logs             Print logs of the job.
+	registry         Add registry secret on paddlecloud.
 	submit           Submit job to PaddlePaddle Cloud.
 
+Subcommands for PFS:
+	cp               uoload or download files
+	ls               List files on PaddlePaddle Cloud
+	mkdir            mkdir directoies on PaddlePaddle Cloud
+	rm               rm files on PaddlePaddle Cloud
+
 
 Use "paddlecloud flags" for a list of top-level flags
 ```
@@ -123,13 +132,30 @@ scp -r my_training_package/ user@tunnel-server:/mnt/hdfs_mulan/idl/idl-dl/mypack
 - 提交基于V1 API的训练任务
 
 ```bash
-paddlecloud submit -jobname my-paddlecloud-job -cpu 1 -gpu 0 -memory 1Gi -parallelism 10 -pscpu 1 -pservers 3 -psmemory 1Gi -passes 1 -topology trainer_config.py /pfs/[datacenter_name]/home/[username]/ctr_demo_package
+paddlecloud submit -jobname my-paddlecloud-job \
+  -cpu 1 \
+  -gpu 0 \
+  -memory 1Gi \
+  -parallelism 10 \
+  -pscpu 1 \
+  -pservers 3 \
+  -psmemory 1Gi \
+  -passes 1 \
+  -topology trainer_config.py /pfs/[datacenter_name]/home/[username]/ctr_demo_package
 ```
 
 - 提交基于V2 API的训练任务
 
 ```bash
-paddlecloud submit -jobname my-paddlecloud-job -cpu 1 -gpu 0 -memory 1Gi -parallelism 10 -pscpu 1 -pservers 3 -psmemory 1Gi -passes 1 -entry "python trainer_config.py" /pfs/[datacenter_name]/home/[username]/ctr_demo_package
+paddlecloud submit -jobname my-paddlecloud-job \
+  -cpu 1 \
+  -gpu 0 \
+  -memory 1Gi \
+  -parallelism 10 \
+  -pscpu 1 \
+  -pservers 3 \
+  -psmemory 1Gi \
+  -entry "python trainer_config.py" /pfs/[datacenter_name]/home/[username]/ctr_demo_package
 ```
 
 参数说明：
@@ -146,6 +172,43 @@ paddlecloud submit -jobname my-paddlecloud-job -cpu 1 -gpu 0 -memory 1Gi -parall
 - `-passes`：执行训练的pass个数
 - `package`：HDFS 训练任务package的路径
 
+### 使用自定义的Runtime Docker Image
+runtime Docker Image是实际被Kubernetes调度的Docker Image，如果在某些情况下需要自定义属于某个任务的Docker Image可以通过以下方式
+- 自定义Runtime Docker Image
+  ```bash
+  git clone https://github.com/PaddlePaddle/cloud.git && cd cloud/docker
+  ./build_docker.sh {PaddlePaddle production image} {runtime Docker image}
+  docker push {runtime Docker image}
+  ```
+- 使用自定义的runtime Docker Image来运行Job
+  ```bash
+  paddlecloud submit -image {runtime Docker image} -jobname ...
+  ```
+
+- 使用私有registry的runtime Docker image
+  - 在PaddleCloud上添加registry认证信息
+    ```bash
+    paddlecloud registry \
+      -username {your username}
+      -password {your password}
+      -server {your registry server}
+      -name {your registry name}
+    ```
+  - 使用私有registry提交任务
+    ```bash
+    paddlecloud submit \
+      -image {runtime Docker image} \
+      -registry {your registry name}
+    ```
+  - 查看所有的registry
+    ```bash
+    paddlecloud get registry
+    ```
+  - 删除指定的registry
+    ```bash
+    paddlecloud delete registry
+    ```
+
 ## 查看任务状态
 
 用户可以查看任务、任务节点、用户空间配额的当前状态。

docker/build_docker.sh

@@ -25,7 +25,7 @@ docker run --rm -it -v $PWD:/cloud $base_image \
 #Build Docker Image
 cat > Dockerfile <<EOF
 FROM ${base_image}
-RUN pip install -U kubernetes && apt-get install -y iputils-ping
+RUN pip install -U kubernetes && apt-get update -y && apt-get install -y iputils-ping
 ADD ./paddle_k8s /usr/bin
 ADD ./k8s_tools.py /root/
 ADD ./python/dist/pcloud-0.1.1-py2-none-any.whl /tmp/

go/cmd/paddlecloud/paddlecloud.go

@@ -18,10 +18,12 @@ func main() {
 	subcommands.Register(&paddlecloud.GetCommand{}, "")
 	subcommands.Register(&paddlecloud.KillCommand{}, "")
 	subcommands.Register(&paddlecloud.SimpleFileCmd{}, "")
-	subcommands.Register(&pfsmod.LsCmd{}, "")
-	subcommands.Register(&pfsmod.CpCmd{}, "")
-	subcommands.Register(&pfsmod.RmCmd{}, "")
-	subcommands.Register(&pfsmod.MkdirCmd{}, "")
+	subcommands.Register(&paddlecloud.RegistryCmd{}, "")
+	subcommands.Register(&paddlecloud.DeleteCommand{}, "")
+	subcommands.Register(&pfsmod.LsCmd{}, "PFS")
+	subcommands.Register(&pfsmod.CpCmd{}, "PFS")
+	subcommands.Register(&pfsmod.RmCmd{}, "PFS")
+	subcommands.Register(&pfsmod.MkdirCmd{}, "PFS")
 
 	flag.Parse()
 	ctx := context.Background()

go/paddlecloud/delete.go

@@ -0,0 +1,52 @@
+package paddlecloud
+
+import (
+	"context"
+	"flag"
+	"fmt"
+	"os"
+
+	"github.com/google/subcommands"
+)
+
+// DeleteCommand do job killings
+type DeleteCommand struct {
+}
+
+// Name is subcommands name
+func (*DeleteCommand) Name() string { return "delete" }
+
+// Synopsis is subcommands synopsis
+func (*DeleteCommand) Synopsis() string { return "Delete the specify resource." }
+
+// Usage is subcommands usage
+func (*DeleteCommand) Usage() string {
+	return `delete registry [registry-name]
+`
+}
+
+// SetFlags registers subcommands flags
+func (p *DeleteCommand) SetFlags(f *flag.FlagSet) {
+}
+
+// Execute kill command
+func (p *DeleteCommand) Execute(_ context.Context, f *flag.FlagSet, _ ...interface{}) subcommands.ExitStatus {
+	if f.NArg() != 2 {
+		f.Usage()
+		return subcommands.ExitFailure
+	}
+	if f.Arg(0) == RegistryCmdName {
+		name := f.Arg(1)
+		r := RegistryCmd{SecretName: KubeRegistryName(name)}
+		err := r.Delete()
+		if err != nil {
+			fmt.Fprintf(os.Stderr, "error delete registry: %v\n", err)
+			return subcommands.ExitFailure
+		}
+		fmt.Fprintf(os.Stdout, "registry: [%s] is deleted\n", name)
+	} else {
+		f.Usage()
+		return subcommands.ExitFailure
+	}
+	return subcommands.ExitSuccess
+}

go/paddlecloud/get.go

@@ -27,7 +27,7 @@ func (*GetCommand) Synopsis() string { return "Print resources" }
 
 // Usage is subcommands usage
 func (*GetCommand) Usage() string {
-	return `get [jobs|workers [jobname]|quota]:
+	return `get [jobs|workers|registry [jobname]|quota]:
 	Print resources.
 `
 }
@@ -47,6 +47,8 @@ func (p *GetCommand) Execute(_ context.Context, f *flag.FlagSet, _ ...interface{
 		jobs()
 	} else if f.Arg(0) == "quota" {
 		quota()
+	} else if f.Arg(0) == "registry" {
+		registry()
 	} else if f.Arg(0) == "workers" {
 		if f.NArg() != 2 {
 			f.Usage()
@@ -91,7 +93,35 @@ func workers(jobname string) error {
 	w.Flush()
 	return nil
 }
-
+func registry() error {
+	respBody, err := utils.GetCall(utils.Config.ActiveConfig.Endpoint+"/api/v1/registry/", nil)
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "err getting registry secret: %v\n", err)
+		return err
+	}
+	var respObj interface{}
+	err = json.Unmarshal(respBody, &respObj)
+	if err != nil {
+		return err
+	}
+	items := respObj.(map[string]interface{})["msg"].(map[string]interface{})["items"].([]interface{})
+	w := tabwriter.NewWriter(os.Stdout, 0, 0, 3, ' ', 0)
+	if len(items) >= 0 {
+		fmt.Fprintf(w, "ID\tNAME\tDATA\n")
+	}
+	idx := 0
+	for _, r := range items {
+		metadata := r.(map[string]interface{})["metadata"].(map[string]interface{})
+		name := RegistryName(metadata["name"].(string))
+		if len(name) != 0 {
+			cTime := metadata["creation_timestamp"].(string)
+			fmt.Fprintf(w, "%d\t%s\t%s\n", idx, name, cTime)
+			idx++
+		}
+	}
+	w.Flush()
+	return err
+}
 func jobs() error {
 	respBody, err := utils.GetCall(utils.Config.ActiveConfig.Endpoint+"/api/v1/jobs/", nil)
 	if err != nil {

go/paddlecloud/registry.go

@@ -0,0 +1,121 @@
+package paddlecloud
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"flag"
+	"fmt"
+	"os"
+	"strings"
+
+	"github.com/PaddlePaddle/cloud/go/utils"
+	"github.com/golang/glog"
+	"github.com/google/subcommands"
+)
+
+const (
+	// RegistryCmdName is subcommand name
+	RegistryCmdName = "registry"
+	// RegistryPrefix is the prefix for Kubernetes secret name
+	RegistryPrefix = "pcloud-registry"
+)
+
+// RegistryCmd is Docker registry secret information
+type RegistryCmd struct {
+	SecretName string `json:"name"`
+	Username   string `json:"username"`
+	Password   string `json:"password"`
+	Server     string `json:"server"`
+}
+
+// Name is the subcommand name
+func (r *RegistryCmd) Name() string { return RegistryCmdName }
+
+// Synopsis is the subcommand's synopsis
+func (r *RegistryCmd) Synopsis() string { return "Add registry secret on paddlecloud." }
+
+// Usage is the subcommand's usage
+func (r *RegistryCmd) Usage() string {
+	return `registry <options> [add|del]:
+`
+}
+
+// SetFlags registers subcommands flags.
+func (r *RegistryCmd) SetFlags(f *flag.FlagSet) {
+	f.StringVar(&r.SecretName, "name", "", "registry secret name")
+	f.StringVar(&r.Username, "username", "", "your Docker registry username")
+	f.StringVar(&r.Password, "password", "", "your Docker registry password")
+	f.StringVar(&r.Server, "server", "", "your Docker registry Server")
+}
+func (r *RegistryCmd) addRegistrySecret() error {
+	jsonString, err := json.Marshal(r)
+	if err != nil {
+		return err
+	}
+	glog.V(10).Infof("Add registry secret: %s to %s\n", jsonString, utils.Config.ActiveConfig.Endpoint+"/api/v1/registry/")
+	respBody, err := utils.PostCall(utils.Config.ActiveConfig.Endpoint+"/api/v1/registry/", jsonString)
+	if err != nil {
+		return err
+	}
+	var respObj interface{}
+	if err = json.Unmarshal(respBody, &respObj); err != nil {
+		return err
+	}
+	// FIXME: Return an error if error message is not empty. Use response code instead
+	errMsg := respObj.(map[string]interface{})["msg"].(string)
+	if len(errMsg) > 0 {
+		return errors.New(errMsg)
+	}
+	return nil
+}
+
+// Delete the specify registry
+func (r *RegistryCmd) Delete() error {
+	jsonString, err := json.Marshal(r)
+	if err != nil {
+		return err
+	}
+	glog.V(10).Infof("Delete registry secret: %s to %s\n", jsonString, utils.Config.ActiveConfig.Endpoint+"/api/v1/registry/")
+	respBody, err := utils.DeleteCall(utils.Config.ActiveConfig.Endpoint+"/api/v1/registry/", jsonString)
+	if err != nil {
+		return err
+	}
+
+	var respObj interface{}
+	if err = json.Unmarshal(respBody, &respObj); err != nil {
+		return err
+	}
+	// FIXME: Return an error if error message is not empty. Use response code instead
+	errMsg := respObj.(map[string]interface{})["msg"].(string)
+	if len(errMsg) > 0 {
+		return errors.New(errMsg)
+	}
+	return nil
+}
+func (r *RegistryCmd) Execute(_ context.Context, f *flag.FlagSet, _ ...interface{}) subcommands.ExitStatus {
+	if r.SecretName == "" || r.Username == "" || r.Password == "" || r.Server == "" {
+		f.Usage()
+		return subcommands.ExitFailure
+	}
+	r.SecretName = strings.Join([]string{RegistryPrefix, r.SecretName}, "-")
+	err := r.addRegistrySecret()
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "add registry secret failed: %s\n", err)
+		return subcommands.ExitFailure
+	}
+	return subcommands.ExitSuccess
+}
+
+// KubeRegistryName add a prefix for the name
+func KubeRegistryName(name string) string {
+	return RegistryPrefix + "-" + name
+}
+
+// RegistryName is registry secret name for PaddleCloud
+func RegistryName(name string) string {
+	if strings.HasPrefix(name, RegistryPrefix) {
+		return name[len(RegistryPrefix)+1 : len(name)]
+	}
+	return ""
+}

go/paddlecloud/submit.go

@@ -30,6 +30,8 @@ type SubmitCmd struct {
 	Topology    string `json:"topology"`
 	Datacenter  string `json:"datacenter"`
 	Passes      int    `json:"passes"`
+	Image       string `json:"image"`
+	Registry    string `json:"registry"`
 }
 
 // Name is subcommands name.
@@ -59,6 +61,8 @@ func (p *SubmitCmd) SetFlags(f *flag.FlagSet) {
 	f.StringVar(&p.Entry, "entry", "", "Command of starting trainer process. Defaults to paddle train")
 	f.StringVar(&p.Topology, "topology", "", "Will Be Deprecated .py file contains paddle v1 job configs")
 	f.IntVar(&p.Passes, "passes", 1, "Pass count for training job")
+	f.StringVar(&p.Image, "image", "", "Runtime Docker image for the job")
+	f.StringVar(&p.Registry, "registry", "", "Registry secret name for the runtime Docker image")
 }
 
 // Execute submit command.

paddlecloud/paddlecloud/urls.py

@@ -32,6 +32,7 @@
     url(r"^api/v1/workers/", paddlejob.views.WorkersView.as_view()),
     url(r"^api/v1/quota/", paddlejob.views.QuotaView.as_view()),
     url(r"^api/v1/file/", paddlejob.views.SimpleFileView.as_view()),
+    url(r"^api/v1/registry/", paddlejob.registry.RegistryView.as_view()),
 ]
 
 urlpatterns += static(settings.MEDIA_URL, document_root=settings.MEDIA_ROOT)

paddlecloud/paddlejob/__init__.py

@@ -1,2 +1,3 @@
 from paddle_job import PaddleJob
-__all__ = ["PaddleJob"]
+import registry
+__all__ = ["PaddleJob", "registry"]