diff --git a/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt b/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt index bbb6066e17d..445c50f8410 100644 --- a/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt +++ b/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt @@ -911,6 +911,9 @@ Release 2.8.0 - UNRELEASED HDFS-7116. Add a command to get the balancer bandwidth (Rakesh R via vinayakumarb) + HDFS-8974. Convert docs in xdoc format to markdown. + (Masatake Iwasaki via aajisaka) + OPTIMIZATIONS HDFS-8026. Trace FSOutputSummer#writeChecksumChunks rather than diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/HdfsRollingUpgrade.md b/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/HdfsRollingUpgrade.md new file mode 100644 index 00000000000..5415912b8f4 --- /dev/null +++ b/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/HdfsRollingUpgrade.md @@ -0,0 +1,293 @@ + + +HDFS Rolling Upgrade +==================== + +* [Introduction](#Introduction) +* [Upgrade](#Upgrade) + * [Upgrade without Downtime](#Upgrade_without_Downtime) + * [Upgrading Non-Federated Clusters](#Upgrading_Non-Federated_Clusters) + * [Upgrading Federated Clusters](#Upgrading_Federated_Clusters) + * [Upgrade with Downtime](#Upgrade_with_Downtime) + * [Upgrading Non-HA Clusters](#Upgrading_Non-HA_Clusters) +* [Downgrade and Rollback](#Downgrade_and_Rollback) +* [Downgrade](#Downgrade) +* [Rollback](#Rollback) +* [Commands and Startup Options for Rolling Upgrade](#Commands_and_Startup_Options_for_Rolling_Upgrade) + * [DFSAdmin Commands](#DFSAdmin_Commands) + * [dfsadmin -rollingUpgrade](#dfsadmin_-rollingUpgrade) + * [dfsadmin -getDatanodeInfo](#dfsadmin_-getDatanodeInfo) + * [dfsadmin -shutdownDatanode](#dfsadmin_-shutdownDatanode) + * [NameNode Startup Options](#NameNode_Startup_Options) + * [namenode -rollingUpgrade](#namenode_-rollingUpgrade) + + +Introduction +------------ + +*HDFS rolling upgrade* allows upgrading individual HDFS daemons. +For examples, the datanodes can be upgraded independent of the namenodes. +A namenode can be upgraded independent of the other namenodes. +The namenodes can be upgraded independent of datanods and journal nodes. + + +Upgrade +------- + +In Hadoop v2, HDFS supports highly-available (HA) namenode services and wire compatibility. +These two capabilities make it feasible to upgrade HDFS without incurring HDFS downtime. +In order to upgrade a HDFS cluster without downtime, the cluster must be setup with HA. + +If there is any new feature which is enabled in new software release, may not work with old software release after upgrade. +In such cases upgrade should be done by following steps. + +1. Disable new feature. +2. Upgrade the cluster. +3. Enable the new feature. + +Note that rolling upgrade is supported only from Hadoop-2.4.0 onwards. + + +### Upgrade without Downtime + +In a HA cluster, there are two or more *NameNodes (NNs)*, many *DataNodes (DNs)*, +a few *JournalNodes (JNs)* and a few *ZooKeeperNodes (ZKNs)*. +*JNs* is relatively stable and does not require upgrade when upgrading HDFS in most of the cases. +In the rolling upgrade procedure described here, +only *NNs* and *DNs* are considered but *JNs* and *ZKNs* are not. +Upgrading *JNs* and *ZKNs* may incur cluster downtime. + +#### Upgrading Non-Federated Clusters + +Suppose there are two namenodes *NN1* and *NN2*, +where *NN1* and *NN2* are respectively in active and standby states. +The following are the steps for upgrading a HA cluster: + +1. Prepare Rolling Upgrade + 1. Run "[`hdfs dfsadmin -rollingUpgrade prepare`](#dfsadmin_-rollingUpgrade)" + to create a fsimage for rollback. + 1. Run "[`hdfs dfsadmin -rollingUpgrade query`](#dfsadmin_-rollingUpgrade)" + to check the status of the rollback image. + Wait and re-run the command until + the "`Proceed with rolling upgrade`" message is shown. +1. Upgrade Active and Standby *NNs* + 1. Shutdown and upgrade *NN2*. + 1. Start *NN2* as standby with the + "[`-rollingUpgrade started`](#namenode_-rollingUpgrade)" option. + 1. Failover from *NN1* to *NN2* + so that *NN2* becomes active and *NN1* becomes standby. + 1. Shutdown and upgrade *NN1*. + 1. Start *NN1* as standby with the + "[`-rollingUpgrade started`](#namenode_-rollingUpgrade)" option. +1. Upgrade *DNs* + 1. Choose a small subset of datanodes (e.g. all datanodes under a particular rack). + 1. Run "[`hdfs dfsadmin -shutdownDatanode upgrade`](#dfsadmin_-shutdownDatanode)" + to shutdown one of the chosen datanodes. + 1. Run "[`hdfs dfsadmin -getDatanodeInfo `](#dfsadmin_-getDatanodeInfo)" + to check and wait for the datanode to shutdown. + 1. Upgrade and restart the datanode. + 1. Perform the above steps for all the chosen datanodes in the subset in parallel. + 1. Repeat the above steps until all datanodes in the cluster are upgraded. +1. Finalize Rolling Upgrade + 1. Run "[`hdfs dfsadmin -rollingUpgrade finalize`](#dfsadmin_-rollingUpgrade)" + to finalize the rolling upgrade. + + +#### Upgrading Federated Clusters + +In a federated cluster, there are multiple namespaces +and a pair of active and standby *NNs* for each namespace. +The procedure for upgrading a federated cluster is similar to upgrading a non-federated cluster +except that Step 1 and Step 4 are performed on each namespace +and Step 2 is performed on each pair of active and standby *NNs*, i.e. + +1. Prepare Rolling Upgrade for Each Namespace +1. Upgrade Active and Standby *NN* pairs for Each Namespace +1. Upgrade *DNs* +1. Finalize Rolling Upgrade for Each Namespace + + +### Upgrade with Downtime + +For non-HA clusters, +it is impossible to upgrade HDFS without downtime since it requires restarting the namenodes. +However, datanodes can still be upgraded in a rolling manner. + + +#### Upgrading Non-HA Clusters + +In a non-HA cluster, there are a *NameNode (NN)*, a *SecondaryNameNode (SNN)* +and many *DataNodes (DNs)*. +The procedure for upgrading a non-HA cluster is similar to upgrading a HA cluster +except that Step 2 "Upgrade Active and Standby *NNs*" is changed to below: + +* Upgrade *NN* and *SNN* + 1. Shutdown *SNN* + 1. Shutdown and upgrade *NN*. + 1. Start *NN* with the + "[`-rollingUpgrade started`](#namenode_-rollingUpgrade)" option. + 1. Upgrade and restart *SNN* + + +Downgrade and Rollback +---------------------- + +When the upgraded release is undesirable +or, in some unlikely case, the upgrade fails (due to bugs in the newer release), +administrators may choose to downgrade HDFS back to the pre-upgrade release, +or rollback HDFS to the pre-upgrade release and the pre-upgrade state. + +Note that downgrade can be done in a rolling fashion but rollback cannot. +Rollback requires cluster downtime. + +Note also that downgrade and rollback are possible only after a rolling upgrade is started and +before the upgrade is terminated. +An upgrade can be terminated by either finalize, downgrade or rollback. +Therefore, it may not be possible to perform rollback after finalize or downgrade, +or to perform downgrade after finalize. + + +Downgrade +--------- + +*Downgrade* restores the software back to the pre-upgrade release +and preserves the user data. +Suppose time *T* is the rolling upgrade start time and the upgrade is terminated by downgrade. +Then, the files created before or after *T* remain available in HDFS. +The files deleted before or after *T* remain deleted in HDFS. + +A newer release is downgradable to the pre-upgrade release +only if both the namenode layout version and the datenode layout version +are not changed between these two releases. + +In a HA cluster, +when a rolling upgrade from an old software release to a new software release is in progress, +it is possible to downgrade, in a rolling fashion, the upgraded machines back to the old software release. +Same as before, suppose *NN1* and *NN2* are respectively in active and standby states. +Below are the steps for rolling downgrade without downtime: + +1. Downgrade *DNs* + 1. Choose a small subset of datanodes (e.g. all datanodes under a particular rack). + 1. Run "[`hdfs dfsadmin -shutdownDatanode upgrade`](#dfsadmin_-shutdownDatanode)" + to shutdown one of the chosen datanodes. + 1. Run "[`hdfs dfsadmin -getDatanodeInfo `](#dfsadmin_-getDatanodeInfo)" + to check and wait for the datanode to shutdown. + 1. Downgrade and restart the datanode. + 1. Perform the above steps for all the chosen datanodes in the subset in parallel. + 1. Repeat the above steps until all upgraded datanodes in the cluster are downgraded. +1. Downgrade Active and Standby *NNs* + 1. Shutdown and downgrade *NN2*. + 1. Start *NN2* as standby normally. + 1. Failover from *NN1* to *NN2* + so that *NN2* becomes active and *NN1* becomes standby. + 1. Shutdown and upgrade *NN1*. + 1. Start *NN1* as standby normally. +1. Finalize Rolling Downgrade + 1. Run "[`hdfs dfsadmin -rollingUpgrade finalize`](#dfsadmin_-rollingUpgrade)" + to finalize the rolling downgrade. + +Note that the datanodes must be downgraded before downgrading the namenodes +since protocols may be changed in a backward compatible manner but not forward compatible, +i.e. old datanodes can talk to the new namenodes but not vice versa. + + +Rollback +-------- + +*Rollback* restores the software back to the pre-upgrade release +but also reverts the user data back to the pre-upgrade state. +Suppose time *T* is the rolling upgrade start time and the upgrade is terminated by rollback. +The files created before *T* remain available in HDFS but the files created after *T* become unavailable. +The files deleted before *T* remain deleted in HDFS but the files deleted after *T* are restored. + +Rollback from a newer release to the pre-upgrade release is always supported. +However, it cannot be done in a rolling fashion. It requires cluster downtime. +Suppose *NN1* and *NN2* are respectively in active and standby states. +Below are the steps for rollback: + +* Rollback HDFS + 1. Shutdown all *NNs* and *DNs*. + 1. Restore the pre-upgrade release in all machines. + 1. Start *NN1* as Active with the + "[`-rollingUpgrade rollback`](#namenode_-rollingUpgrade)" option. + 1. Run `-bootstrapStandby' on NN2 and start it normally as standby. + 1. Start *DNs* with the "`-rollback`" option. + + +Commands and Startup Options for Rolling Upgrade +------------------------------------------------ + +### DFSAdmin Commands + +#### `dfsadmin -rollingUpgrade` + + hdfs dfsadmin -rollingUpgrade + +Execute a rolling upgrade action. + +* Options: + + | --- | --- | + | `query` | Query the current rolling upgrade status. | + | `prepare` | Prepare a new rolling upgrade. | + | `finalize` | Finalize the current rolling upgrade. | + + +#### `dfsadmin -getDatanodeInfo` + + hdfs dfsadmin -getDatanodeInfo + +Get the information about the given datanode. +This command can be used for checking if a datanode is alive +like the Unix `ping` command. + + +#### `dfsadmin -shutdownDatanode` + + hdfs dfsadmin -shutdownDatanode [upgrade] + +Submit a shutdown request for the given datanode. +If the optional `upgrade` argument is specified, +clients accessing the datanode will be advised to wait for it to restart +and the fast start-up mode will be enabled. +When the restart does not happen in time, clients will timeout and ignore the datanode. +In such case, the fast start-up mode will also be disabled. + +Note that the command does not wait for the datanode shutdown to complete. +The "[`dfsadmin -getDatanodeInfo`](#dfsadmin_-getDatanodeInfo)" +command can be used for checking if the datanode shutdown is completed. + + +### NameNode Startup Options + +#### `namenode -rollingUpgrade` + + hdfs namenode -rollingUpgrade + +When a rolling upgrade is in progress, +the `-rollingUpgrade` namenode startup option is used to specify +various rolling upgrade options. + +* Options: + + | --- | --- | + | `rollback` | Restores the namenode back to the pre-upgrade release but also reverts the user data back to the pre-upgrade state. | + | `started` | Specifies a rolling upgrade already started so that the namenode should allow image directories with different layout versions during startup. | + +**WARN: downgrade options is obsolete.** +It is not necessary to start namenode with downgrade options explicitly. diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/HdfsSnapshots.md b/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/HdfsSnapshots.md new file mode 100644 index 00000000000..94a37cd77c1 --- /dev/null +++ b/hadoop-hdfs-project/hadoop-hdfs/src/site/markdown/HdfsSnapshots.md @@ -0,0 +1,301 @@ + + +HDFS Snapshots +============== + +* [HDFS Snapshots](#HDFS_Snapshots) + * [Overview](#Overview) + * [Snapshottable Directories](#Snapshottable_Directories) + * [Snapshot Paths](#Snapshot_Paths) + * [Upgrading to a version of HDFS with snapshots](#Upgrading_to_a_version_of_HDFS_with_snapshots) + * [Snapshot Operations](#Snapshot_Operations) + * [Administrator Operations](#Administrator_Operations) + * [Allow Snapshots](#Allow_Snapshots) + * [Disallow Snapshots](#Disallow_Snapshots) + * [User Operations](#User_Operations) + * [Create Snapshots](#Create_Snapshots) + * [Delete Snapshots](#Delete_Snapshots) + * [Rename Snapshots](#Rename_Snapshots) + * [Get Snapshottable Directory Listing](#Get_Snapshottable_Directory_Listing) + * [Get Snapshots Difference Report](#Get_Snapshots_Difference_Report) + + +Overview +-------- + +HDFS Snapshots are read-only point-in-time copies of the file system. +Snapshots can be taken on a subtree of the file system or the entire file system. +Some common use cases of snapshots are data backup, protection against user errors +and disaster recovery. + +The implementation of HDFS Snapshots is efficient: + + +* Snapshot creation is instantaneous: + the cost is *O(1)* excluding the inode lookup time. + +* Additional memory is used only when modifications are made relative to a snapshot: + memory usage is *O(M)*, + where *M* is the number of modified files/directories. + +* Blocks in datanodes are not copied: + the snapshot files record the block list and the file size. + There is no data copying. + +* Snapshots do not adversely affect regular HDFS operations: + modifications are recorded in reverse chronological order + so that the current data can be accessed directly. + The snapshot data is computed by subtracting the modifications + from the current data. + + +### Snapshottable Directories + +Snapshots can be taken on any directory once the directory has been set as +*snapshottable*. +A snapshottable directory is able to accommodate 65,536 simultaneous snapshots. +There is no limit on the number of snapshottable directories. +Administrators may set any directory to be snapshottable. +If there are snapshots in a snapshottable directory, +the directory can be neither deleted nor renamed +before all the snapshots are deleted. + +Nested snapshottable directories are currently not allowed. +In other words, a directory cannot be set to snapshottable +if one of its ancestors/descendants is a snapshottable directory. + + +### Snapshot Paths + +For a snapshottable directory, +the path component *".snapshot"* is used for accessing its snapshots. +Suppose `/foo` is a snapshottable directory, +`/foo/bar` is a file/directory in `/foo`, +and `/foo` has a snapshot `s0`. +Then, the path `/foo/.snapshot/s0/bar` +refers to the snapshot copy of `/foo/bar`. +The usual API and CLI can work with the ".snapshot" paths. +The following are some examples. + +* Listing all the snapshots under a snapshottable directory: + + hdfs dfs -ls /foo/.snapshot + +* Listing the files in snapshot `s0`: + + hdfs dfs -ls /foo/.snapshot/s0 + +* Copying a file from snapshot `s0`: + + hdfs dfs -cp -ptopax /foo/.snapshot/s0/bar /tmp + + Note that this example uses the preserve option to preserve + timestamps, ownership, permission, ACLs and XAttrs. + + +Upgrading to a version of HDFS with snapshots +--------------------------------------------- + +The HDFS snapshot feature introduces a new reserved path name used to +interact with snapshots: `.snapshot`. When upgrading from an +older version of HDFS, existing paths named `.snapshot` need +to first be renamed or deleted to avoid conflicting with the reserved path. +See the upgrade section in +[the HDFS user guide](HdfsUserGuide.html#Upgrade_and_Rollback) +for more information. + + +Snapshot Operations +------------------- + + +### Administrator Operations + +The operations described in this section require superuser privilege. + + +#### Allow Snapshots + + +Allowing snapshots of a directory to be created. +If the operation completes successfully, the directory becomes snapshottable. + +* Command: + + hdfs dfsadmin -allowSnapshot + +* Arguments: + + | --- | --- | + | path | The path of the snapshottable directory. | + +See also the corresponding Java API +`void allowSnapshot(Path path)` in `HdfsAdmin`. + + +#### Disallow Snapshots + +Disallowing snapshots of a directory to be created. +All snapshots of the directory must be deleted before disallowing snapshots. + +* Command: + + hdfs dfsadmin -disallowSnapshot + +* Arguments: + + | --- | --- | + | path | The path of the snapshottable directory. | + +See also the corresponding Java API +`void disallowSnapshot(Path path)` in `HdfsAdmin`. + + +### User Operations + +The section describes user operations. +Note that HDFS superuser can perform all the operations +without satisfying the permission requirement in the individual operations. + + +#### Create Snapshots + +Create a snapshot of a snapshottable directory. +This operation requires owner privilege of the snapshottable directory. + +* Command: + + hdfs dfs -createSnapshot [] + +* Arguments: + + | --- | --- | + | path | The path of the snapshottable directory. | + | snapshotName | The snapshot name, which is an optional argument. When it is omitted, a default name is generated using a timestamp with the format `"'s'yyyyMMdd-HHmmss.SSS"`, e.g. `"s20130412-151029.033"`. | + +See also the corresponding Java API +`Path createSnapshot(Path path)` and +`Path createSnapshot(Path path, String snapshotName)` +in [FileSystem](../../api/org/apache/hadoop/fs/FileSystem.html) +The snapshot path is returned in these methods. + + +#### Delete Snapshots + +Delete a snapshot of from a snapshottable directory. +This operation requires owner privilege of the snapshottable directory. + +* Command: + + hdfs dfs -deleteSnapshot + +* Arguments: + + | --- | --- | + | path | The path of the snapshottable directory. | + | snapshotName | The snapshot name. | + +See also the corresponding Java API +`void deleteSnapshot(Path path, String snapshotName)` +in [FileSystem](../../api/org/apache/hadoop/fs/FileSystem.html). + + +#### Rename Snapshots + +Rename a snapshot. +This operation requires owner privilege of the snapshottable directory. + +* Command: + + hdfs dfs -renameSnapshot + +* Arguments: + + | --- | --- | + | path | The path of the snapshottable directory. | + | oldName | The old snapshot name. | + | newName | The new snapshot name. | + +See also the corresponding Java API +`void renameSnapshot(Path path, String oldName, String newName)` +in [FileSystem](../../api/org/apache/hadoop/fs/FileSystem.html). + + +#### Get Snapshottable Directory Listing + +Get all the snapshottable directories where the current user has permission to take snapshtos. + +* Command: + + hdfs lsSnapshottableDir + +* Arguments: none + +See also the corresponding Java API +`SnapshottableDirectoryStatus[] getSnapshottableDirectoryListing()` +in `DistributedFileSystem`. + + +#### Get Snapshots Difference Report + +Get the differences between two snapshots. +This operation requires read access privilege for all files/directories in both snapshots. + +* Command: + + hdfs snapshotDiff + +* Arguments: + + | --- | --- | + | path | The path of the snapshottable directory. | + | fromSnapshot | The name of the starting snapshot. | + | toSnapshot | The name of the ending snapshot. | + + Note that snapshotDiff can be used to get the difference report between two snapshots, or between + a snapshot and the current status of a directory. Users can use "." to represent the current status. + +* Results: + + | --- | --- | + | \+ | The file/directory has been created. | + | \- | The file/directory has been deleted. | + | M | The file/directory has been modified. | + | R | The file/directory has been renamed. | + +A *RENAME* entry indicates a file/directory has been renamed but +is still under the same snapshottable directory. A file/directory is +reported as deleted if it was renamed to outside of the snapshottble directory. +A file/directory renamed from outside of the snapshottble directory is +reported as newly created. + +The snapshot difference report does not guarantee the same operation sequence. +For example, if we rename the directory *"/foo"* to *"/foo2"*, and +then append new data to the file *"/foo2/bar"*, the difference report will +be: + + R. /foo -> /foo2 + M. /foo/bar + +I.e., the changes on the files/directories under a renamed directory is +reported using the original path before the rename (*"/foo/bar"* in +the above example). + +See also the corresponding Java API +`SnapshotDiffReport getSnapshotDiffReport(Path path, String fromSnapshot, String toSnapshot)` +in `DistributedFileSystem`. diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/site/xdoc/HdfsRollingUpgrade.xml b/hadoop-hdfs-project/hadoop-hdfs/src/site/xdoc/HdfsRollingUpgrade.xml deleted file mode 100644 index f0b0ccf0df0..00000000000 --- a/hadoop-hdfs-project/hadoop-hdfs/src/site/xdoc/HdfsRollingUpgrade.xml +++ /dev/null @@ -1,329 +0,0 @@ - - - - - - HDFS Rolling Upgrade - - - - -

HDFS Rolling Upgrade

- - - - - - -
-

- HDFS rolling upgrade allows upgrading individual HDFS daemons. - For examples, the datanodes can be upgraded independent of the namenodes. - A namenode can be upgraded independent of the other namenodes. - The namenodes can be upgraded independent of datanods and journal nodes. -

-
- -
-

- In Hadoop v2, HDFS supports highly-available (HA) namenode services and wire compatibility. - These two capabilities make it feasible to upgrade HDFS without incurring HDFS downtime. - In order to upgrade a HDFS cluster without downtime, the cluster must be setup with HA. -

-

- If there is any new feature which is enabled in new software release, may not work with old software release after upgrade. - In such cases upgrade should be done by following steps. -

-
    -
  1. Disable new feature.
    2. -
  2. Upgrade the cluster.
    3. -
  3. Enable the new feature.
    4. -
-

- Note that rolling upgrade is supported only from Hadoop-2.4.0 onwards. -

- -

- In a HA cluster, there are two or more NameNodes (NNs), many DataNodes (DNs), - a few JournalNodes (JNs) and a few ZooKeeperNodes (ZKNs). - JNs is relatively stable and does not require upgrade when upgrading HDFS in most of the cases. - In the rolling upgrade procedure described here, - only NNs and DNs are considered but JNs and ZKNs are not. - Upgrading JNs and ZKNs may incur cluster downtime. -

- -

Upgrading Non-Federated Clusters

-

- Suppose there are two namenodes NN1 and NN2, - where NN1 and NN2 are respectively in active and standby states. - The following are the steps for upgrading a HA cluster: -

-
    -
  1. Prepare Rolling Upgrade
      -
    1. Run "hdfs dfsadmin -rollingUpgrade prepare" - to create a fsimage for rollback. -
      2. -
    2. Run "hdfs dfsadmin -rollingUpgrade query" - to check the status of the rollback image. - Wait and re-run the command until - the "Proceed with rolling upgrade" message is shown. -
      3. -
    2. -
  2. Upgrade Active and Standby NNs
      -
    1. Shutdown and upgrade NN2.
      2. -
    2. Start NN2 as standby with the - "-rollingUpgrade started" option.
      3. -
    3. Failover from NN1 to NN2 - so that NN2 becomes active and NN1 becomes standby.
      4. -
    4. Shutdown and upgrade NN1.
      5. -
    5. Start NN1 as standby with the - "-rollingUpgrade started" option.
      6. -
    3. -
  3. Upgrade DNs
      -
    1. Choose a small subset of datanodes (e.g. all datanodes under a particular rack).
      2. -
        -
      1. Run "hdfs dfsadmin -shutdownDatanode <DATANODE_HOST:IPC_PORT> upgrade" - to shutdown one of the chosen datanodes.
        2. -
      2. Run "hdfs dfsadmin -getDatanodeInfo <DATANODE_HOST:IPC_PORT>" - to check and wait for the datanode to shutdown.
        3. -
      3. Upgrade and restart the datanode.
        4. -
      4. Perform the above steps for all the chosen datanodes in the subset in parallel.
        5. -
      -
    2. Repeat the above steps until all datanodes in the cluster are upgraded.
      3. -
    4. -
  4. Finalize Rolling Upgrade
    5. -
- -

Upgrading Federated Clusters

-

- In a federated cluster, there are multiple namespaces - and a pair of active and standby NNs for each namespace. - The procedure for upgrading a federated cluster is similar to upgrading a non-federated cluster - except that Step 1 and Step 4 are performed on each namespace - and Step 2 is performed on each pair of active and standby NNs, i.e. -

-
    -
  1. Prepare Rolling Upgrade for Each Namespace
    2. -
  2. Upgrade Active and Standby NN pairs for Each Namespace
    3. -
  3. Upgrade DNs
    4. -
  4. Finalize Rolling Upgrade for Each Namespace
    5. -
- - - - -

- For non-HA clusters, - it is impossible to upgrade HDFS without downtime since it requires restarting the namenodes. - However, datanodes can still be upgraded in a rolling manner. -

- -

Upgrading Non-HA Clusters

-

- In a non-HA cluster, there are a NameNode (NN), a SecondaryNameNode (SNN) - and many DataNodes (DNs). - The procedure for upgrading a non-HA cluster is similar to upgrading a HA cluster - except that Step 2 "Upgrade Active and Standby NNs" is changed to below: -

-
    -
  • Upgrade NN and SNN
      -
    1. Shutdown SNN
      2. -
    2. Shutdown and upgrade NN.
      3. -
    3. Start NN with the - "-rollingUpgrade started" option.
      4. -
    4. Upgrade and restart SNN
      5. -
    • -
- -
- -
-

- When the upgraded release is undesirable - or, in some unlikely case, the upgrade fails (due to bugs in the newer release), - administrators may choose to downgrade HDFS back to the pre-upgrade release, - or rollback HDFS to the pre-upgrade release and the pre-upgrade state. -

-

- Note that downgrade can be done in a rolling fashion but rollback cannot. - Rollback requires cluster downtime. -

-

- Note also that downgrade and rollback are possible only after a rolling upgrade is started and - before the upgrade is terminated. - An upgrade can be terminated by either finalize, downgrade or rollback. - Therefore, it may not be possible to perform rollback after finalize or downgrade, - or to perform downgrade after finalize. -

-
- -
-

- Downgrade restores the software back to the pre-upgrade release - and preserves the user data. - Suppose time T is the rolling upgrade start time and the upgrade is terminated by downgrade. - Then, the files created before or after T remain available in HDFS. - The files deleted before or after T remain deleted in HDFS. -

-

- A newer release is downgradable to the pre-upgrade release - only if both the namenode layout version and the datenode layout version - are not changed between these two releases. -

-

- In a HA cluster, - when a rolling upgrade from an old software release to a new software release is in progress, - it is possible to downgrade, in a rolling fashion, the upgraded machines back to the old software release. - Same as before, suppose NN1 and NN2 are respectively in active and standby states. - Below are the steps for rolling downgrade without downtime: -

-
    -
  1. Downgrade DNs
      -
    1. Choose a small subset of datanodes (e.g. all datanodes under a particular rack).
      2. -
        -
      1. Run "hdfs dfsadmin -shutdownDatanode <DATANODE_HOST:IPC_PORT> upgrade" - to shutdown one of the chosen datanodes.
        2. -
      2. Run "hdfs dfsadmin -getDatanodeInfo <DATANODE_HOST:IPC_PORT>" - to check and wait for the datanode to shutdown.
        3. -
      3. Downgrade and restart the datanode.
        4. -
      4. Perform the above steps for all the chosen datanodes in the subset in parallel.
        5. -
      -
    2. Repeat the above steps until all upgraded datanodes in the cluster are downgraded.
      3. -
    2. -
  2. Downgrade Active and Standby NNs
      -
    1. Shutdown and downgrade NN2.
      2. -
    2. Start NN2 as standby normally. -
      3. -
    3. Failover from NN1 to NN2 - so that NN2 becomes active and NN1 becomes standby.
      4. -
    4. Shutdown and upgrade NN1.
      5. -
    5. Start NN1 as standby normally. -
      6. -
    3. -
  3. Finalize Rolling Downgrade
    4. -
-

- Note that the datanodes must be downgraded before downgrading the namenodes - since protocols may be changed in a backward compatible manner but not forward compatible, - i.e. old datanodes can talk to the new namenodes but not vice versa. -

-
- -
-

- Rollback restores the software back to the pre-upgrade release - but also reverts the user data back to the pre-upgrade state. - Suppose time T is the rolling upgrade start time and the upgrade is terminated by rollback. - The files created before T remain available in HDFS but the files created after T become unavailable. - The files deleted before T remain deleted in HDFS but the files deleted after T are restored. -

-

- Rollback from a newer release to the pre-upgrade release is always supported. - However, it cannot be done in a rolling fashion. It requires cluster downtime. - Suppose NN1 and NN2 are respectively in active and standby states. - Below are the steps for rollback: -

-
    -
  • Rollback HDFS
      -
    1. Shutdown all NNs and DNs.
      2. -
    2. Restore the pre-upgrade release in all machines.
      3. -
    3. Start NN1 as Active with the - "-rollingUpgrade rollback" option.
      4. -
    4. Run `-bootstrapStandby' on NN2 and start it normally as standby.
      5. -
    5. Start DNs with the "-rollback" option.
      6. -
    • -
- -
- -
- - -

dfsadmin -rollingUpgrade

- hdfs dfsadmin -rollingUpgrade <query|prepare|finalize> -

- Execute a rolling upgrade action. -

  • Options: - - - -
    queryQuery the current rolling upgrade status.
    preparePrepare a new rolling upgrade.
    finalizeFinalize the current rolling upgrade.
-

- -

dfsadmin -getDatanodeInfo

- hdfs dfsadmin -getDatanodeInfo <DATANODE_HOST:IPC_PORT> -

- Get the information about the given datanode. - This command can be used for checking if a datanode is alive - like the Unix ping command. -

- -

dfsadmin -shutdownDatanode

- hdfs dfsadmin -shutdownDatanode <DATANODE_HOST:IPC_PORT> [upgrade] -

- Submit a shutdown request for the given datanode. - If the optional upgrade argument is specified, - clients accessing the datanode will be advised to wait for it to restart - and the fast start-up mode will be enabled. - When the restart does not happen in time, clients will timeout and ignore the datanode. - In such case, the fast start-up mode will also be disabled. -

-

- Note that the command does not wait for the datanode shutdown to complete. - The "dfsadmin -getDatanodeInfo" - command can be used for checking if the datanode shutdown is completed. -

- - - - -

namenode -rollingUpgrade

- hdfs namenode -rollingUpgrade <rollback|started> -

- When a rolling upgrade is in progress, - the -rollingUpgrade namenode startup option is used to specify - various rolling upgrade options. -

-
  • Options: - - - - - - -
    rollbackRestores the namenode back to the pre-upgrade release - but also reverts the user data back to the pre-upgrade state.
    startedSpecifies a rolling upgrade already started - so that the namenode should allow image directories - with different layout versions during startup.
-

- WARN: downgrade options is obsolete. - It is not necessary to start namenode with downgrade options explicitly. -

- - -
- - diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/site/xdoc/HdfsSnapshots.xml b/hadoop-hdfs-project/hadoop-hdfs/src/site/xdoc/HdfsSnapshots.xml deleted file mode 100644 index 330d00fdef1..00000000000 --- a/hadoop-hdfs-project/hadoop-hdfs/src/site/xdoc/HdfsSnapshots.xml +++ /dev/null @@ -1,303 +0,0 @@ - - - - - - HDFS Snapshots - - - - -

HDFS Snapshots

- - - - - - -
-

- HDFS Snapshots are read-only point-in-time copies of the file system. - Snapshots can be taken on a subtree of the file system or the entire file system. - Some common use cases of snapshots are data backup, protection against user errors - and disaster recovery. -

- -

- The implementation of HDFS Snapshots is efficient: -

-
    -
  • Snapshot creation is instantaneous: - the cost is O(1) excluding the inode lookup time.
    • -
  • Additional memory is used only when modifications are made relative to a snapshot: - memory usage is O(M), - where M is the number of modified files/directories.
    • -
  • Blocks in datanodes are not copied: - the snapshot files record the block list and the file size. - There is no data copying.
    • -
  • Snapshots do not adversely affect regular HDFS operations: - modifications are recorded in reverse chronological order - so that the current data can be accessed directly. - The snapshot data is computed by subtracting the modifications - from the current data.
    • -
- - -

- Snapshots can be taken on any directory once the directory has been set as - snapshottable. - A snapshottable directory is able to accommodate 65,536 simultaneous snapshots. - There is no limit on the number of snapshottable directories. - Administrators may set any directory to be snapshottable. - If there are snapshots in a snapshottable directory, - the directory can be neither deleted nor renamed - before all the snapshots are deleted. -

- -

- Nested snapshottable directories are currently not allowed. - In other words, a directory cannot be set to snapshottable - if one of its ancestors/descendants is a snapshottable directory. -

- - - - -

- For a snapshottable directory, - the path component ".snapshot" is used for accessing its snapshots. - Suppose /foo is a snapshottable directory, - /foo/bar is a file/directory in /foo, - and /foo has a snapshot s0. - Then, the path /foo/.snapshot/s0/bar - refers to the snapshot copy of /foo/bar. - The usual API and CLI can work with the ".snapshot" paths. - The following are some examples. -

-
    -
  • Listing all the snapshots under a snapshottable directory: - hdfs dfs -ls /foo/.snapshot
    • -
  • Listing the files in snapshot s0: - hdfs dfs -ls /foo/.snapshot/s0
    • -
  • Copying a file from snapshot s0: - hdfs dfs -cp -ptopax /foo/.snapshot/s0/bar /tmp -

    Note that this example uses the preserve option to preserve - timestamps, ownership, permission, ACLs and XAttrs.

    • -
- -
- -
- -

- The HDFS snapshot feature introduces a new reserved path name used to - interact with snapshots: .snapshot. When upgrading from an - older version of HDFS, existing paths named .snapshot need - to first be renamed or deleted to avoid conflicting with the reserved path. - See the upgrade section in - the HDFS user guide - for more information.

- -
- -
- -

- The operations described in this section require superuser privilege. -

- -

Allow Snapshots

-

- Allowing snapshots of a directory to be created. - If the operation completes successfully, the directory becomes snapshottable. -

-
    -
  • Command: - hdfs dfsadmin -allowSnapshot <path>
    • -
  • Arguments: - -
    pathThe path of the snapshottable directory.
    • -
-

- See also the corresponding Java API - void allowSnapshot(Path path) in HdfsAdmin. -

- -

Disallow Snapshots

-

- Disallowing snapshots of a directory to be created. - All snapshots of the directory must be deleted before disallowing snapshots. -

-
    -
  • Command: - hdfs dfsadmin -disallowSnapshot <path>
    • -
  • Arguments: - -
    pathThe path of the snapshottable directory.
    • -
-

- See also the corresponding Java API - void disallowSnapshot(Path path) in HdfsAdmin. -

- - - -

- The section describes user operations. - Note that HDFS superuser can perform all the operations - without satisfying the permission requirement in the individual operations. -

- -

Create Snapshots

-

- Create a snapshot of a snapshottable directory. - This operation requires owner privilege of the snapshottable directory. -

-
    -
  • Command: - hdfs dfs -createSnapshot <path> [<snapshotName>]
    • -
  • Arguments: - - -
    pathThe path of the snapshottable directory.
    snapshotName - The snapshot name, which is an optional argument. - When it is omitted, a default name is generated using a timestamp with the format - "'s'yyyyMMdd-HHmmss.SSS", e.g. "s20130412-151029.033". -
    • -
-

- See also the corresponding Java API - Path createSnapshot(Path path) and - Path createSnapshot(Path path, String snapshotName) - in FileSystem. - The snapshot path is returned in these methods. -

- -

Delete Snapshots

-

- Delete a snapshot of from a snapshottable directory. - This operation requires owner privilege of the snapshottable directory. -

-
    -
  • Command: - hdfs dfs -deleteSnapshot <path> <snapshotName>
    • -
  • Arguments: - - -
    pathThe path of the snapshottable directory.
    snapshotNameThe snapshot name.
    • -
-

- See also the corresponding Java API - void deleteSnapshot(Path path, String snapshotName) - in FileSystem. -

- -

Rename Snapshots

-

- Rename a snapshot. - This operation requires owner privilege of the snapshottable directory. -

-
    -
  • Command: - hdfs dfs -renameSnapshot <path> <oldName> <newName>
    • -
  • Arguments: - - - -
    pathThe path of the snapshottable directory.
    oldNameThe old snapshot name.
    newNameThe new snapshot name.
    • -
-

- See also the corresponding Java API - void renameSnapshot(Path path, String oldName, String newName) - in FileSystem. -

- -

Get Snapshottable Directory Listing

-

- Get all the snapshottable directories where the current user has permission to take snapshtos. -

-
    -
  • Command: - hdfs lsSnapshottableDir
    • -
  • Arguments: none
    • -
-

- See also the corresponding Java API - SnapshottableDirectoryStatus[] getSnapshottableDirectoryListing() - in DistributedFileSystem. -

- -

Get Snapshots Difference Report

-

- Get the differences between two snapshots. - This operation requires read access privilege for all files/directories in both snapshots. -

-
    -
  • Command: - hdfs snapshotDiff <path> <fromSnapshot> <toSnapshot>
    • -
  • Arguments: - - - -
    pathThe path of the snapshottable directory.
    fromSnapshotThe name of the starting snapshot.
    toSnapshotThe name of the ending snapshot.
    • -

    - Note that snapshotDiff can be used to get the difference report between two snapshots, or between - a snapshot and the current status of a directory.Users can use "." to represent the current status. -

    -
  • Results: - - - - - -
    +The file/directory has been created.
    -The file/directory has been deleted.
    MThe file/directory has been modified.
    RThe file/directory has been renamed.
    -
    • -
-

- A RENAME entry indicates a file/directory has been renamed but - is still under the same snapshottable directory. A file/directory is - reported as deleted if it was renamed to outside of the snapshottble directory. - A file/directory renamed from outside of the snapshottble directory is - reported as newly created. -

-

- The snapshot difference report does not guarantee the same operation sequence. - For example, if we rename the directory "/foo" to "/foo2", and - then append new data to the file "/foo2/bar", the difference report will - be: - - R. /foo -> /foo2 - M. /foo/bar - - I.e., the changes on the files/directories under a renamed directory is - reported using the original path before the rename ("/foo/bar" in - the above example). -

-

- See also the corresponding Java API - SnapshotDiffReport getSnapshotDiffReport(Path path, String fromSnapshot, String toSnapshot) - in DistributedFileSystem. -

- - -
- - -