HDFS
Last updated
Last updated
This guide describes the instructions to configure as Alluxio's under storage system.
To run an Alluxio cluster on a set of machines, you must deploy Alluxio server binaries to each of these machines. You can either with the correct Hadoop version (recommended), or (for advanced users).
Note that, when building Alluxio from source code, by default Alluxio server binaries are built to work with Apache Hadoop HDFS of version 3.3.4. To work with Hadoop distributions of other versions, one needs to specify the correct Hadoop profile and run the following in your Alluxio directory:
Alluxio provides predefined build profiles for hadoop-2 and hadoop-3 (enabled by default) for the major Hadoop versions 2.x and 3.x. If you want to build Alluxio with a specific Hadoop release version, you can also specify the version in the command. For example,
Please visit the page for more information about support for other distributions.
If everything succeeds, you should see alluxio-assembly-server-{{site.ALLUXIO_VERSION_STRING}}-jar-with-dependencies.jar created in the ${ALLUXIO_HOME}/assembly/server/target directory.
To enable active sync be sure to build using the hdfsActiveSync property. For example,
Please visit for more information on using active sync.
To configure Alluxio to use HDFS as under storage, you will need to modify the configuration file conf/alluxio-site.properties. If the file does not exist, create the configuration file from the template.
Edit conf/alluxio-site.properties file to set the under storage address to the HDFS namenode address and the HDFS directory you want to mount to Alluxio. For example, the under storage address can be hdfs://localhost:8020 if you are running the HDFS namenode locally with default port and mapping HDFS root directory to Alluxio, or hdfs://localhost:8020/alluxio/data if only the HDFS directory /alluxio/data is mapped to Alluxio. To find out where HDFS is running, use hdfs getconf -confKey fs.defaultFS to get the default hostname and port HDFS is listening on.
Before this step, make sure your HDFS cluster is running and the directory mapped to Alluxio exists. Start the Alluxio servers:
If your ramdisk is not mounted, the format command can fail. This is likely because this is the first time you are running Alluxio, you may need to start Alluxio with the SudoMount option.
Run a simple example program:
If the test fails with permission errors, make sure that the current user (${USER}) has read/write access to the HDFS directory mounted to Alluxio. By default, the login user is the current user of the host OS. To change the user, set the value of alluxio.security.login.username in conf/alluxio-site.properties to the desired username.
Stop Alluxio by running:
When HDFS has non-default configurations, you need to configure Alluxio servers to access HDFS with the proper configuration file. Note that once this is set, your applications using Alluxio client do not need any special configuration.
There are two possible approaches:
Copy or make symbolic links from hdfs-site.xml and core-site.xml from your Hadoop installation into ${ALLUXIO_HOME}/conf. Make sure this is set up on all servers running Alluxio.
Alternatively, you can set the property alluxio.master.mount.table.root.option.alluxio.underfs.hdfs.configuration in conf/alluxio-site.properties to point to your hdfs-site.xml and core-site.xml. Make sure this configuration is set on all servers running Alluxio.
The user set above is only the identity that starts Alluxio master and worker processes. Once Alluxio servers are started, it is unnecessary to run your Alluxio client applications using this user.
In addition, security configuration is needed for Alluxio to be able to communicate with the HDFS cluster. Set the following Alluxio properties in alluxio-site.properties:
If connecting to secure HDFS, run kinit on all Alluxio nodes. Use the principal hdfs and the keytab that you configured earlier in alluxio-site.properties. A known limitation is that the Kerberos TGT may expire after the max renewal lifetime. You can work around this by renewing the TGT periodically. Otherwise you may see No valid credentials provided (Mechanism level: Failed to find any Kerberos tgt) when starting Alluxio services. Another option is to set alluxio.hadoop.security.kerberos.keytab.login.autorenewal=true so the TGT is automatically refreshed.
The user can also use alluxio.hadoop.security.krb5.conf to specify the krb5.conf file location and use alluxio.hadoop.security.authentication to specify authentication method.
By default, Alluxio will use machine-level Kerberos configuration to determine the Kerberos realm and KDC. You can override these defaults by setting the JVM properties java.security.krb5.realm and java.security.krb5.kdc.
To set these, set ALLUXIO_JAVA_OPTS in conf/alluxio-env.sh.
There are multiple ways for a user to mount an HDFS cluster with a specified version as an under storage into Alluxio namespace. Before mounting HDFS with a specific version, make sure you have built a client with that specific version of HDFS. You can check the existence of this client by going to the lib directory under the Alluxio directory.
If you have built Alluxio from source, you can build additional client jar files by running mvn command under the underfs directory in the Alluxio source tree. For example, issuing the following command would build the client jar for the 2.8.0 version.
When using the mount Alluxio shell command, one can pass through the mount option alluxio.underfs.version to specify which version of HDFS to mount. If no such a version is specified, by default Alluxio treats it as Apache HDFS 3.3.
For example, the following commands mount two HDFS deployments—one is HDFS 2.2 and the other is 2.7—into Alluxio namespace under directory /mnt/hdfs22 and /mnt/hdfs27.
When mounting the under storage of Alluxio root directory with a specific HDFS version, one can add the following line to the site properties file (conf/alluxio-site.properties)
Alluxio supports the following versions of HDFS as a valid argument of mount option alluxio.underfs.version:
Apache Hadoop: 2.2, 2.3, 2.4, 2.5, 2.6, 2.7, 2.8, 2.9, 2.10, 3.0, 3.1, 3.2, 3.3
Make sure to restart Alluxio services for the change to take effect.
Additionally, you may need to specify the following property to be your HDFS version. See .
This will start one Alluxio master and one Alluxio worker locally. You can see the master UI at .
After this succeeds, you can visit HDFS web UI at ( if you are running HDFS 2.x) to verify the files and directories created by Alluxio exist. For this test, you should see files named like: /default_tests_files/BASIC_CACHE_THROUGH at
To configure Alluxio to work with HDFS namenodes in HA mode, first configure Alluxio servers to .
In addition, set the under storage address to hdfs://nameservice/ (nameservice is the already configured in hdfs-site.xml). To mount an HDFS subdirectory to Alluxio instead of the whole HDFS namespace, change the under storage address to something like hdfs://nameservice/alluxio/data.
Alluxio supports POSIX-like filesystem . To ensure that the permission information of files/directories including user, group and mode in HDFS is consistent with Alluxio (e.g., a file created by user Foo in Alluxio is persisted to HDFS also with owner as user Foo), the user to start Alluxio master and worker processes is required to be either:
. Namely, use the same user that starts HDFS namenode process to also start Alluxio master and worker processes.
A member of . Edit HDFS configuration file hdfs-site.xml and check the value of configuration property dfs.permissions.superusergroup. If this property is set with a group (e.g., "hdfs"), add the user to start Alluxio process (e.g., "alluxio") to this group ("hdfs"); if this property is not set, add a group to this property where your Alluxio running user is a member of this newly added group.
If your HDFS cluster is Kerberized, first configure Alluxio servers to .
Note: Apache Hadoop 1.0 and 1.2 are still supported, but not included in the default download. To build this module yourself, build the shaded hadoop client and then the UFS module. Please refer to instructions to see how to build additional UFS libraries.
Hadoop comes with a native library that provides better performance and additional features compared to its Java implementation. For example, when the native library is used, the HDFS client can use native checksum function which is more efficient than the default Java implementation. To use the Hadoop native library with Alluxio HDFS under filesystem, first install the native library on Alluxio nodes by following the instructions on . Once the hadoop native library is installed on the machine, update Alluxio startup Java parameters in conf/alluxio-env.sh by adding the following line: