Skip to content

Files

Latest commit

 

History

History
 
 

repository-hdfs

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
May 8, 2017

⚠️ Starting with ES-Hadoop 5.x, the plugin has moved to Elasticsearch

The HDFS Snapshot/Restore plugin has been moved into Elasticsearch proper. When upgrading, please keep in mind that the light and 1.x versions are no longer supported in 5.x.

More information can be found here. For feature requests or bugs regarding the Repository HDFS plugin, please open an issue in the Elasticsearch repository with the ":Plugin Repository HDFS" tag.

ES-Hadoop 2.x instructions below

elasticsearch-repository-hdfs plugin allows Elasticsearch 2.0 to use hdfs file-system as a repository for snapshot/restore. See this blog entry for a quick introduction to snapshot/restore.

Requirements

  • Elasticsearch (version 2.x).
  • HDFS accessible file-system (from the Elasticsearch classpath)
  • Elasticsearch Java Security Manager disabled (due to permission issues with HDFS)

Flavors

The HDFS snapshot/restore plugin comes in three flavors:

  • Default / Hadoop 1.x The default version contains the plugin jar alongside Hadoop 1.x (stable) dependencies
  • Yarn / Hadoop 2.x The hadoop2 version contains the plugin jar plus the Hadoop 2.x (Yarn) dependencies.
  • Light The light version contains just the plugin jar, without any Hadoop dependencies.

What version to use?

It depends on whether you have Hadoop installed on your nodes or not. If you do, then it may be easiest to expose Hadoop to the Elasticsearch classpath (typically by creating symbolic links to the client jars within the Elasticsearch lib/ directory) and using the light version. This guarantees the existing libraries and configuration are being picked up by the plugin. If you do not have Hadoop installed, then select either the default version (for Hadoop stable/1.x) or, if you are using Hadoop 2, the hadoop2 version.

Installation

The HDFS Snapshot/Restore is an Elasticsearch plugin - be sure to familiarize with what these are and how they work by reading the plugins chapter in the Elasticsearch documentation.

Disable the Java Security Manager (JSM)

By default, Elasticsearch enforces a Java Security Manager inside its running JVM for security purposes. Unfortunately Hadoop and its HDFS client are quite greedy in terms of the permissions needed, requiring not just significantly more permissions than Elasticsearch itself but also dangerous ones. Thus, one is required to disable the JSM, otherwise the plugin will not work correctly; this can be done by adding security.manager.enabled: false to the elasticsearch.yml configuration on each node where the plugin runs.

One can easily check whether the JSM is disabled or not by looking at the logs for this warning:

[2015-10-25 23:13:45,478][INFO ][plugin.hadoop.hdfs       ] Loaded Hadoop [1.2.1] libraries from file:/xxx/plugins/repository-hdfs/
[2015-10-25 23:13:45,478][WARN ][plugin.hadoop.hdfs       ] The Java Security Manager is enabled; unfortunately Hadoop is not compatible with it so it needs to be disabled; see the docs for more information...

If the warning appears, the JSM is enabled. If it does not (after the message indicating the libraries have been loaded) then everything is fine.

Wait, why can't the JSM be used?

Security is hard.

While efforts like these help with per-plugin permissions, the ultimate goal is having a secure Elasticsearch install. Unfortunately Hadoop (especially 2.x) requires dangerous permissions such as execute on all files (triggered during even basic initialization. Permissions like this (not to mention the shell execution) are simply too dangerous and security is significantly affected. Where possible, we try to find the common ground and have the code still running securely with hacks like these.

As the above is addressed only in master (potentially the upcoming Elasticsearch 2.3), in the meantime users need to be aware of the current situation and act accordingly. That is, understand that by using HDFS plugin the Java Security Manager is disabled.

Node restart

After installing the plugin on every Elasticsearch node, be sure to restart it. This applies to all nodes on which the plugins have been installed - without restarting the nodes, the plugin will not function properly.

Stable version

As with any other plugin, simply run: bin/plugin install elasticsearch/elasticsearch-repository-hdfs/2.2.0

When looking for light or hadoop2 artifacts use: bin/plugin install elasticsearch/elasticsearch-repository-hdfs/2.2.0-<classifier>

Development Snapshot

To install the latest snapshot, please install the plugin manually using: bin/plugin install <url-path-to-plugin.zip>

Or grab the latest nightly build from the repository again through Maven:

<dependency>
  <groupId>org.elasticsearch</groupId>
  <artifactId>elasticsearch-repository-hdfs</artifactId>
  <version>2.3.0.BUILD-SNAPSHOT</version>
</dependency>
<repositories>
  <repository>
    <id>sonatype-oss</id>
    <url>http://oss.sonatype.org/content/repositories/snapshots</url>
    <snapshots><enabled>true</enabled></snapshots>
  </repository>
</repositories>

Configuration Properties

Once installed, define the configuration for the hdfs repository through elasticsearch.yml or the REST API:

repositories
  hdfs:
    uri: "hdfs://<host>:<port>/"    # optional - Hadoop file-system URI
    path: "some/path"               # required - path with the file-system where data is stored/loaded
    load_defaults: "true"           # optional - whether to load the default Hadoop configuration (default) or not
    conf_location: "extra-cfg.xml"  # optional - Hadoop configuration XML to be loaded (use commas for multi values)
    conf.<key> : "<value>"          # optional - 'inlined' key=value added to the Hadoop configuration
    concurrent_streams: 5           # optional - the number of concurrent streams (defaults to 5)
    compress: "false"               # optional - whether to compress the metadata or not (default)
    chunk_size: "10mb"              # optional - chunk size (disabled by default)

NOTE: Be careful when including a paths within the uri setting; Some implementations ignore them completely while others consider them. In general, we recommend keeping the uri to a minimum and using the path element instead.

Plugging other file-systems

Any HDFS-compatible file-systems (like Amazon s3:// or Google gs://) can be used as long as the proper Hadoop configuration is passed to the Elasticsearch plugin. In practice, this means making sure the correct Hadoop configuration files (core-site.xml and hdfs-site.xml) and its jars are available in plugin classpath, just as you would with any other Hadoop client or job. Otherwise, the plugin will only read the default, vanilla configuration of Hadoop and will not be able to recognized the plugged in file-system.

Feedback / Q&A

We're interested in your feedback! You can find us on the forum - please use the Hadoop channel. For more details, see the community page.

License

This project is released under version 2.0 of the Apache License

Licensed to Elasticsearch under one or more contributor
license agreements. See the NOTICE file distributed with
this work for additional information regarding copyright
ownership. Elasticsearch licenses this file to you under
the Apache License, Version 2.0 (the "License"); you may
not use this file except in compliance with the License.
You may obtain a copy of the License at

   http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing,
software distributed under the License is distributed on an
"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
KIND, either express or implied.  See the License for the
specific language governing permissions and limitations
under the License.