1. HDFS快照管理

快照相当于对目录做一个备份。并不会立即复制所有文件，而是记录文件变化

1.1. 开启快照功能

开启指定目录的快照功能

1 2	$ hdfs dfsadmin -allowSnapshot /input Allowing snaphot on /input succeeded

1.2. 创建快照

-createSnapshot <snapshotDir> [<snapshotName>] 创建一个快照

1 2	$ hdfs dfs -createSnapshot /input Created snapshot /input/.snapshot/s20190225-152050.714

默认是看不到隐藏文件的

$ hdfs dfs -ls -R /input     # 看不到.snapshot目录，实际是存在的
-rw-r--r--   3 root supergroup        280 2019-02-25 14:55 /input/1.txt
-rw-r--r--   3 root supergroup         58 2019-02-25 15:17 /input/2.txt
-rw-r--r--   3 root supergroup          4 2019-02-25 15:17 /input/3.txt
-rw-r--r--   3 root supergroup          4 2019-02-25 15:18 /input/4.txt

$ hdfs dfs -ls -R /input/.snapshot   # 直接查看
drwxr-xr-x   - root supergroup          0 2019-02-25 15:20 /input/.snapshot/s20190225-152050.714
-rw-r--r--   3 root supergroup        280 2019-02-25 14:55 /input/.snapshot/s20190225-152050.714/1.txt
-rw-r--r--   3 root supergroup         58 2019-02-25 15:17 /input/.snapshot/s20190225-152050.714/2.txt
-rw-r--r--   3 root supergroup          4 2019-02-25 15:17 /input/.snapshot/s20190225-152050.714/3.txt
-rw-r--r--   3 root supergroup          4 2019-02-25 15:18 /input/.snapshot/s20190225-152050.714/4.txt

删除一个文件

1	hdfs dfs -rm -r -f /input/1.txt

再创建一个快照，可指定名称

1 2	$ hdfs dfs -createSnapshot /input snapshot2 Created snapshot /input/.snapshot/snapshot2

1.3. 列出有快照的目录

列出当前用户有快照的目录

1 2	$ hdfs lsSnapshottableDir drwxr-xr-x 0 root supergroup 0 2019-02-25 15:20 1 65536 /input

1.4. 重命名快照

-renameSnapshot <snapshotDir> <oldName> <newName>

1 2	hdfs dfs -renameSnapshot /input snapshot2 snapshot3 # snapshot2重命名为snapshot3 hdfs dfs -renameSnapshot /input snapshot3 snapshot2 # 恢复修改

1.5. 比较快照不同以及目录变化

snapshotDiff用法：hdfs snapshotDiff <snapshotDir> <from> <to>

<snapshotDir>：有快照的目录
<from/to>：
- 可以是 . ，表示该目录的当前状态
- 也可以是.snapshot/snapshot_name，代表该目录的某个快照。其中.snapshot可以省略，简写为snapshot_name

比较两个快照的不同

$ hdfs snapshotDiff /input s20190225-152050.714 snapshot2
Difference between snapshot s20190225-152050.714 and snapshot snapshot2 under directory /input:
M       .          # 当前目录发生变化
-       ./1.txt    # 后者比前者少了1.txt

$ hdfs snapshotDiff /input snapshot2 s20190225-152050.714  # 参数的一下
Difference between snapshot snapshot2 and snapshot s20190225-152050.714 under directory /input:
M       .
+       ./1.txt      # 后者比前者多了1.txt

比较当前状态与快照的不同

$ hdfs dfs -rm -r -f /input/2.txt
$ hdfs dfs -put input/5.txt /input/
$ hdfs dfs -rm -r -f /input/3.txt
$ hdfs dfs -put input/3.txt /input
$ hdfs dfs -appendToFile - /input/4.txt
hello world
^D   # CTRL+D

$ hdfs snapshotDiff /input snapshot2 .    # 后者比前者
Difference between snapshot snapshot2 and current directory under directory /input:
M       .
+       ./3.txt
+       ./5.txt
-       ./2.txt
-       ./3.txt
M       ./4.txt

$ hdfs snapshotDiff /input . snapshot2    # 后者比前者
Difference between current directory and snapshot snapshot2 under directory /input:
M       .
-       ./3.txt
-       ./5.txt
+       ./2.txt
+       ./3.txt
M       ./4.txt

1.6. 恢复快照

这些的恢复快照不能一次性回到原来的状态，只能手动拷贝。将快照目录复制出来即可

$ hdfs dfs -cp /input/.snapshot/snapshot2 /   # 复制snapshot2目录
$ hdfs dfs -ls -R /snapshot2
-rw-r--r--   3 root supergroup         58 2019-02-25 16:13 /snapshot2/2.txt
-rw-r--r--   3 root supergroup          4 2019-02-25 16:13 /snapshot2/3.txt
-rw-r--r--   3 root supergroup          4 2019-02-25 16:13 /snapshot2/4.txt

1.7. 删除快照

-deleteSnapshot <snapshotDir> <snapshotName>

1 2	hdfs dfs -deleteSnapshot /input s20190225-152050.714 hdfs dfs -deleteSnapshot /input snapshot2

1.8. 禁用快照

hdfs dfsadmin [-disallowSnapshot <snapshotDir>]

$ hdfs dfs -createSnapshot /input snapshot2       # 先创建一个快照
Created snapshot /input/.snapshot/snapshot2

$ hdfs dfsadmin -disallowSnapshot /input    # 尝试禁用，失败，如果当前目录有快照，则无法禁止
disallowSnapshot: The directory /input has snapshot(s). Please redo the operation after removing all the snapshots.

$ hdfs dfs -deleteSnapshot /input snapshot2  # 先删除该目录的所有快照

$ hdfs dfsadmin -disallowSnapshot /input  # 再禁用
Disallowing snaphot on /input succeeded