add the function of restoring from metadata been backuped (fluid-clou…

…dnative#564) * add the function of restoring from metadata been backuped * code improve Co-authored-by: 仇伶玮 <[email protected]>
shiwenwenya · Feb 9, 2021 · 90cdc4c · 90cdc4c
1 parent a56e352
commit 90cdc4c
Show file tree

Hide file tree

Showing 13 changed files with 357 additions and 53 deletions.
diff --git a/api/v1alpha1/dataset_types.go b/api/v1alpha1/dataset_types.go
@@ -100,6 +100,16 @@ type Mount struct {
 	EncryptOptions []EncryptOption `json:"encryptOptions,omitempty"`
 }
 
+// DataRestoreLocation describes the spec restore location of  Dataset
+type DataRestoreLocation struct {
+	// Path describes the path of restore, in the form of  local://subpath or pvc://<pvcName>/subpath
+	// +optional
+	Path string `json:"path,omitempty"`
+	// NodeName describes the nodeName of restore if Path is  in the form of local://subpath
+	// +optional
+	NodeName string `json:"nodeName,omitempty"`
+}
+
 // DatasetSpec defines the desired state of Dataset
 type DatasetSpec struct {
 	// Mount Points to be mounted on Alluxio.
@@ -133,6 +143,10 @@ type DatasetSpec struct {
 	// +kubebuilder:validation:Enum=Exclusive;"";Shared
 	// +optional
 	PlacementMode PlacementMode `json:"placement,omitempty"`
+
+	// DataRestoreLocation is the location to load data of dataset  been backuped
+	// +optional
+	DataRestoreLocation *DataRestoreLocation `json:"dataRestoreLocation,omitempty"`
 }
 
 // Runtime describes a runtime to be used to support dataset

diff --git a/api/v1alpha1/zz_generated.deepcopy.go b/api/v1alpha1/zz_generated.deepcopy.go
diff --git a/charts/alluxio/templates/master/statefulset.yaml b/charts/alluxio/templates/master/statefulset.yaml
@@ -158,6 +158,15 @@ spec:
             name: embedded
           {{- end }}
           volumeMounts:
+          {{ if .Values.master.restore.enabled -}}
+          {{- if .Values.master.restore.pvcName }}
+          - name: pvc
+            mountPath: /pvc
+          {{- else }}
+          - name: host
+            mountPath: /host
+        {{- end }}
+        {{- end }}
           {{- if .Values.master.backupPath }}
           - name: backup
             mountPath: /alluxio_backups
@@ -292,6 +301,18 @@ spec:
           {{- end }}
       restartPolicy: Always
       volumes:
+        {{ if .Values.master.restore.enabled -}}
+        {{- if .Values.master.restore.pvcName }}
+        - name: pvc
+          persistentVolumeClaim:
+            claimName: {{ .Values.master.restore.pvcName }}
+        {{- else }}
+        - name: host
+          hostPath:
+            path: {{ .Values.master.restore.path }}
+            type: DirectoryOrCreate
+        {{- end }}
+        {{- end }}
         {{- if .Values.master.backupPath }}
         - name: backup
           hostPath:

diff --git a/charts/alluxio/values.yaml b/charts/alluxio/values.yaml
@@ -106,7 +106,13 @@ master:
   # JVM options specific to the master container
   jvmOptions:
   nodeSelector: {}
-  backupPath:
+  # backupPath is the path in the host to save metadata and   metadata info
+  backupPath: #<backupPath>
+  # restore is the location to restore metadata and metadata info  from
+  restore:
+    enabled: false
+    # path: /subpath/
+    # pvcName: pvc-local
 
 jobMaster:
   args:

diff --git a/charts/fluid/fluid/crds/data.fluid.io_datasets.yaml b/charts/fluid/fluid/crds/data.fluid.io_datasets.yaml
@@ -73,6 +73,19 @@ spec:
               items:
                 type: string
               type: array
+            dataRestoreLocation:
+              description: DataRestoreLocation is the location to load data of dataset  been
+                backuped
+              properties:
+                nodeName:
+                  description: NodeName describes the nodeName of restore if Path
+                    is  in the form of local://subpath
+                  type: string
+                path:
+                  description: Path describes the path of restore, in the form of  local://subpath
+                    or pvc://<pvcName>/subpath
+                  type: string
+              type: object
             mounts:
               description: Mount Points to be mounted on Alluxio.
               items:

diff --git a/config/crd/bases/data.fluid.io_datasets.yaml b/config/crd/bases/data.fluid.io_datasets.yaml
@@ -73,6 +73,19 @@ spec:
               items:
                 type: string
               type: array
+            dataRestoreLocation:
+              description: DataRestoreLocation is the location to load data of dataset  been
+                backuped
+              properties:
+                nodeName:
+                  description: NodeName describes the nodeName of restore if Path
+                    is  in the form of local://subpath
+                  type: string
+                path:
+                  description: Path describes the path of restore, in the form of  local://subpath
+                    or pvc://<pvcName>/subpath
+                  type: string
+              type: object
             mounts:
               description: Mount Points to be mounted on Alluxio.
               items:

diff --git a/docs/zh/samples/backup_and_restore_metadata.md b/docs/zh/samples/backup_and_restore_metadata.md
@@ -114,4 +114,101 @@ $ kubectl describe node cn-beijing.192.168.1.146
 获得IP地址后，就可以从主机的对应目录取走文件刚刚的两个文件
 
 
-## 恢复
+## 恢复
+要进行恢复，需要保证Dataset的名称与原来保持一致，否则会找不到备份文件
+
+### 从PVC恢复
+在创建Dataset时，在spec中添加dataRestoreLocation，填入刚刚查询到的路径作为恢复路径
+
+如果备份文件被移动过，需要修改路径
+```bash
+$ cat <<EOF > dataset.yaml
+apiVersion: data.fluid.io/v1alpha1
+kind: Dataset
+metadata:
+  name: hbase
+spec:
+  dataRestoreLocation:
+    path: pvc://pvc-local/subpath1/
+  mounts:
+    - mountPoint:  https://mirrors.tuna.tsinghua.edu.cn/apache/hbase/2.2.6/
+EOF
+```
+
+创建Dataset资源对象：
+```bash
+$ kubectl create -f dataset.yaml
+dataset.data.fluid.io/hbase created
+```
+
+AlluxioRuntime资源对象无需做特殊配置：
+```bash
+$ cat<<EOF >runtime.yaml
+apiVersion: data.fluid.io/v1alpha1
+kind: AlluxioRuntime
+metadata:
+  name: hbase
+spec:
+  replicas: 2
+  tieredstore:
+    levels:
+      - mediumtype: MEM
+        path: /dev/shm
+        quota: 2Gi
+        high: "0.95"
+        low: "0.7"
+  properties:
+    alluxio.user.block.size.bytes.default: 256MB
+    alluxio.user.streaming.reader.chunk.size.bytes: 256MB
+    alluxio.user.local.reader.chunk.size.bytes: 256MB
+    alluxio.worker.network.reader.buffer.size: 256MB
+    alluxio.user.streaming.data.timeout: 300sec
+  fuse:
+    args:
+      - fuse
+      -  --fuse-opts=kernel_cache,ro,max_read=131072,attr_timeout=7200,entry_timeout=7200,nonempty,max_readahead=0
+EOF
+```
+创建AlluxioRuntime资源对象：
+```bash
+$ kubectl create -f runtime.yaml
+alluxioruntime.data.fluid.io/hbase created
+```
+数据集加载时，将不再从UFS中加载metadata并统计UFS TOTAL SIZE等信息，而是从备份文件中恢复
+
+片刻后，Dataset进入Bound状态：
+```bash
+$ kubectl get dataset
+NAME    UFS TOTAL SIZE   CACHED   CACHE CAPACITY   CACHED  PERCENTAGE   PHASE   AGE
+hbase   443.86MiB        0.00B    4.00GiB          0.0%                 Bound   20h
+```
+### 从本地恢复
+在创建Dataset时，在spec中添加dataRestoreLocation，填入刚刚查询到的路径和备份文件所在主机的nodeName
+
+如果备份文件被移动过，需要修改路径和nodeName
+```bash
+$ cat <<EOF > dataset.yaml
+apiVersion: data.fluid.io/v1alpha1
+kind: Dataset
+metadata:
+  name: hbase
+spec:
+  dataRestoreLocation:
+    path: local:///data/subpath1/
+    nodeName: cn-beijing.192.168.1.146
+  mounts:
+    - mountPoint:  https://mirrors.tuna.tsinghua.edu.cn/apache/hbase/2.2.6/
+EOF
+```
+创建AlluxioRuntime资源对象：
+```bash
+$ kubectl create -f runtime.yaml
+alluxioruntime.data.fluid.io/hbase created
+```
+片刻后，Dataset进入Bound状态
+```bash
+$ kubectl get dataset
+NAME    UFS TOTAL SIZE   CACHED   CACHE CAPACITY   CACHED  PERCENTAGE   PHASE   AGE
+hbase   443.86MiB        0.00B    4.00GiB          0.0%                 Bound   20h
+```
+
diff --git a/pkg/ddc/alluxio/dataset.go b/pkg/ddc/alluxio/dataset.go
@@ -17,7 +17,6 @@ package alluxio
 
 import (
 	"context"
-	"github.com/fluid-cloudnative/fluid/pkg/ddc/alluxio/operations"
 	"reflect"
 
 	datav1alpha1 "github.com/fluid-cloudnative/fluid/api/v1alpha1"
@@ -127,22 +126,6 @@ func (e *AlluxioEngine) UpdateDatasetStatus(phase datav1alpha1.DatasetPhase) (er
 		} else {
 			e.Log.Info("No need to update HCFS status")
 		}
-
-		newFileNum, err := e.getDataSetFileNum()
-		if err != nil {
-			return err
-		} else if newFileNum != datasetToUpdate.Status.FileNum {
-			datasetToUpdate.Status.FileNum = newFileNum
-			// backup the filenum result in local
-			podName, containerName := e.getMasterPodInfo()
-			fileUtils := operations.NewAlluxioFileUtils(podName, containerName, e.namespace, e.Log)
-			metadataInfoFile := e.GetMetadataInfoFile()
-			err = fileUtils.InsertMetaDataInfoIntoFile(operations.FileNum, newFileNum, metadataInfoFile)
-			if err != nil {
-				e.Log.Error(err, "Failed to InsertMetaDataInfoIntoFile of the dataset")
-			}
-		}
-
 		e.Log.Info("the dataset status", "status", datasetToUpdate.Status)
 
 		if !reflect.DeepEqual(dataset.Status, datasetToUpdate.Status) {