HDFS file system support#

Trino includes support to access the Hadoop Distributed File System (HDFS) with a catalog using the Delta Lake, Hive, Hudi, or Iceberg connectors.

Support is enabled by default, but can be deactivated by setting fs.hadoop.enabled to false.

Apache Hadoop HDFS 2.x and 3.x are supported.

General configuration#

Use the following properties to configure general aspects of HDFS support:




Activate the support for HDFS access. Defaults to true. Must be set to true for all other properties be used.


An optional, comma-separated list of HDFS configuration files. These files must exist on the machines running Trino. For basic setups, Trino configures the HDFS client automatically and does not require any configuration files. In some cases, such as when using federated HDFS or NameNode high availability, it is necessary to specify additional HDFS client options to access your HDFS cluster in the HDFS XML configuration files and reference them with this parameter:


Only specify additional configuration files if necessary for your setup, and reduce the configuration files to have the minimum set of required properties. Additional properties may cause problems.


Controls the permissions set on new directories created for schemas and tables. Value must either be skip or an octal number, with a leading 0. If set to skip, permissions of newly created directories are not set by Trino. Defaults to 0777.


Flag to determine if new files inherit the ownership information from the directory. Defaults to false.


Flag to determine if file checksums must be verified. Defaults to false.


Duration between IPC pings from Trino to HDFS. Defaults to 10s.


Timeout duration for access operations on HDFS. Defaults to 60s.


Timeout duration for connection operations to HDFS. Defaults to 500ms.


Maximum number of retries for HDFS connection attempts. Defaults to 5.


Caching time duration for the key provider. Defaults to 30min.


Path to the UNIX domain socket for the DataNode. The path must exist on each node. For example, /var/lib/hadoop-hdfs/dn_socket.


URL for a SOCKS proxy to use for accessing HDFS. For example, hdfs-master:1180.


Enable HDFS wire encryption. In a Kerberized Hadoop cluster that uses HDFS wire encryption, this must be set to true to enable Trino to access HDFS. Note that using wire encryption may impact query execution performance. Defaults to false.


Maximum number of cached file system objects in the HDFS cache. Defaults to 1000.


Integer value to set the HDFS replication factor. By default, no value is set.


HDFS support includes capabilities for user impersonation and Kerberos authentication. The following properties are available:

Property value



Configure the authentication to use no authentication (NONE) or Kerberos authentication (KERBEROS). Defaults to NONE.


Enable HDFS end-user impersonation. Defaults to false. See details in HDFS impersonation.


The Kerberos principal Trino uses when connecting to HDFS. Example: trino-hdfs-superuser/trino-server-node@EXAMPLE.COM or trino-hdfs-superuser/_HOST@EXAMPLE.COM.

The _HOST placeholder can be used in this property value. When connecting to HDFS, the Hive connector substitutes in the hostname of the worker node Trino is running on. This is useful if each worker node has its own Kerberos principal.


The path to the keytab file that contains a key for the principal specified by hive.hdfs.trino.principal. This file must be readable by the operating system user running Trino.


The location of the credential-cachewiuth the credentials for the principal to use to access HDFS. Altenative to hive.hdfs.trino.keytab.

The default security configuration does not use authentication when connecting to a Hadoop cluster (hive.hdfs.authentication.type=NONE). All queries are executed as the OS user who runs the Trino process, regardless of which user submits the query.

Before running any CREATE TABLE or CREATE TABLE AS statements for Hive tables in Trino, you must check that the user Trino is using to access HDFS has access to the Hive warehouse directory. The Hive warehouse directory is specified by the configuration variable hive.metastore.warehouse.dir in hive-site.xml, and the default value is /user/hive/warehouse.

For example, if Trino is running as nobody, it accesses HDFS as nobody. You can override this username by setting the HADOOP_USER_NAME system property in the Trino JVM config, replacing hdfs_user with the appropriate username:


The hive user generally works, since Hive is often started with the hive user and this user has access to the Hive warehouse.

HDFS impersonation#

HDFS impersonation is enabled by adding hive.hdfs.impersonation.enabled=true to the catalog properties file. With this configuration HDFS, Trino can impersonate the end user who is running the query. This can be used with HDFS permissions and ACLs to provide additional security for data. HDFS permissions and ACLs are explained in the HDFS Permissions Guide.

To use impersonation, the Hadoop cluster must be configured to allow the user or principal that Trino is running as to impersonate the users who log in to Trino. Impersonation in Hadoop is configured in the file core-site.xml. A complete description of the configuration options is available in the Hadoop documentation.

In the case of a user running a query from the command line interface, the end user is the username associated with the Trino CLI process or argument to the optional --user option.

HDFS Kerberos authentication#

To use Trino with a Hadoop cluster that uses Kerberos authentication, you must configure the catalog in the catalog properties file to work with two services on the Hadoop cluster:

Both setups require that Kerberos is configured on each Trino node. Access to the Trino coordinator must be secured, for example using Kerberos or password authentication, when using Kerberos authentication to Hadoop services. Failure to secure access to the Trino coordinator could result in unauthorized access to sensitive data on the Hadoop cluster. Refer to Security for further information, and specifically consider configuring Kerberos authentication.


If your krb5.conf location is different from /etc/krb5.conf you must set it explicitly using the java.security.krb5.conf JVM property in the jvm.config file. For example, -Djava.security.krb5.conf=/example/path/krb5.conf.

Keytab files#

Keytab files are needed for Kerberos authentication and contain encryption keys that are used to authenticate principals to the Kerberos KDC. These encryption keys must be stored securely; you must take the same precautions to protect them that you take to protect ssh private keys.

In particular, access to keytab files must be limited to only the accounts that must use them to authenticate. In practice, this is the user that the Trino process runs as. The ownership and permissions on keytab files must be set to prevent other users from reading or modifying the files.

Keytab files must be distributed to every node running Trino, and must have the correct permissions on every node after distributing them.

Security configuration examples#

The following sections describe the configuration properties and values needed for the various authentication configurations with HDFS.

Default NONE authentication without impersonation#


The default authentication type for HDFS is NONE. When the authentication type is NONE, Trino connects to HDFS using Hadoop’s simple authentication mechanism. Kerberos is not used.

NONE authentication with impersonation#


When using NONE authentication with impersonation, Trino impersonates the user who is running the query when accessing HDFS. The user Trino is running as must be allowed to impersonate this user, as discussed in the section HDFS impersonation. Kerberos is not used.

KERBEROS authentication without impersonation#


When the authentication type is KERBEROS, Trino accesses HDFS as the principal specified by the hive.hdfs.trino.principal property. Trino authenticates this principal using the keytab specified by the hive.hdfs.trino.keytab keytab.

KERBEROS authentication with impersonation#


When using KERBEROS authentication with impersonation, Trino impersonates the user who is running the query when accessing HDFS. The principal specified by the hive.hdfs.trino.principal property must be allowed to impersonate the current Trino user, as discussed in the section HDFS impersonation. Trino authenticates hive.hdfs.trino.principal using the keytab specified by hive.hdfs.trino.keytab.