8.7. Presto-Admin Commands

catalog add

presto-admin catalog add [<name>]

This command is used to deploy catalog configurations to the Presto cluster. Catalog configurations are kept in the configuration directory ``~/.prestoadmin/catalog`

To add a catalog using presto-admin, first create a configuration file in ~/.prestoadmin/catalog. The file should be named <name>.properties and contain the configuration for that catalog.

Use the optional name argument to add a particular catalog to your cluster. To deploy all catalogs in the catalog configuration directory, leave the name argument out.

In order to query using the newly added catalog, you need to restart the Presto server (see server restart):

presto-admin server restart

Example

To add a catalog for the jmx connector, create a file ~/.prestoadmin/catalog/jmx.properties with the content connector.name=jmx. Then run:

./presto-admin catalog add jmx
./presto-admin server restart

If you have two catalog configurations in the catalog directory, for example jmx.properties and dummy.properties, and would like to deploy both at once, you could run

./presto-admin catalog add
./presto-admin server restart

Adding a Custom Connector

In order to install a catalog for a custom connector not included with Presto, the jar must be added to the Presto plugin location using the plugin add_jar command before running the catalog add command.

Example:

./presto-admin plugin add_jar my_connector.jar my_connector
./presto-admin catalog add my_connector
./presto-admin server restart

The add_jar command assumes the default plugin location of /usr/lib/presto/lib/plugin (see plugin add_jar). As with the default connectors, a my_connector.properties file must be created. Refer to the custom connector’s documentation for the properties to specify.

The plugin add_jar command works with both jars and directories containing jars.

catalog remove

presto-admin catalog remove <name>

The catalog remove command is used to remove a catalog from your presto cluster configuration. Running the command will remove the catalog from all nodes in the Presto cluster. Additionally, it will remove the local configuration file for the catalog.

In order for the change to take effect, you will need to restart services.

presto-admin server restart

Example

For example: To remove the catalog for the jmx connector, run

./presto-admin catalog remove jmx
./presto-admin server restart

collect logs

presto-admin collect logs

This command gathers Presto server logs and launcher logs from the /var/log/presto/ directory across the cluster along with the ~/.prestoadmin/log/presto-admin.log and creates a tar file. The final tar output will be saved at /tmp/presto-debug-logs.tar.bz2.

Example

./presto-admin collect logs

collect query_info

presto-admin collect query_info <query_id>

This command gathers information about a Presto query identified by the given query_id and stores that information in a JSON file. The output file will be saved at /tmp/presto-debug/query_info_<query_id>.json.

Example

./presto-admin collect query_info 20150525_234711_00000_7qwaz

collect system_info

presto-admin collect system_info

This command gathers various system specific information from the cluster. The information is saved in a tar file at /tmp/presto-debug-sysinfo.tar.bz2. The gathered information includes:

  • Node specific information from Presto like node uri, last response time, recent failures, recent requests made to the node, etc.
  • List of catalogs configured
  • Catalog configuration files
  • Other system specific information like OS information, Java version, presto-admin version and Presto server version

Example

./presto-admin collect system_info

configuration deploy

presto-admin configuration deploy [coordinator|workers]

This command deploys onto the cluster. presto-admin uses different configuration directories for worker and coordinator configurations so that you can easily create different configurations for your coordinator and worker nodes. Create a ~/.prestoadmin/coordinator directory for your coordinator configurations and a ~/.prestoadmin/workers directory for your workers configuration. If you have the presto-admin configuration directory path set using the environment variable PRESTO_ADMIN_CONFIG_DIR then the coordinator and worker configuration directories must be created under $PRESTO_ADMIN_CONFIG_DIR. Place the configuration files for the coordinator and workers in their respective directories. The optional coordinator or workers argument tells presto-admin to only deploy the coordinator or workers configurations. To deploy both configurations at once, don’t specify either option.

When you run configuration deploy, the following files will be deployed to the /etc/presto directory on your Presto cluster:

  • node.properties
  • config.properties
  • jvm.config
  • log.properties (if it exists)

Note

This command will not deploy the configurations for catalogs. To deploy catalog configurations run catalog add

If the coordinator is also a worker, it will get the coordinator configuration. The deployed configuration files will overwrite the existing configurations on the cluster. However, the node.id from the node.properties file will be preserved. If no node.id exists, a new id will be generated. If any required files are absent when you run configuration deploy, a default configuration will be deployed. Below are the default configurations:

node.properties

node.environment=presto
node.data-dir=/var/lib/presto/data
node.launcher-log-file=/var/log/presto/launcher.log
node.server-log-file=/var/log/presto/server.log
catalog.config-dir=/etc/presto/catalog
plugin.dir=/usr/lib/presto/lib/plugin

Note

Do not change the value of catalog.config-dir=/etc/presto/catalog as it is necessary for Presto to be able to find the catalog directory when Presto has been installed by RPM.

jvm.config

-server
-Xmx16G
-XX:-UseBiasedLocking
-XX:+UseG1GC
-XX:G1HeapRegionSize=32M
-XX:+ExplicitGCInvokesConcurrent
-XX:+HeapDumpOnOutOfMemoryError
-XX:+UseGCOverheadLimit
-XX:OnOutOfMemoryError=kill -9 %p
-XX:ReservedCodeCacheSize=512M
-DHADOOP_USER_NAME=hive

config.properties

For workers:

coordinator=false
discovery.uri=http://<coordinator>:8080
http-server.http.port=8080
query.max-memory-per-node=8GB
query.max-memory=50GB

For coordinator:

coordinator=true
discovery-server.enabled=true
discovery.uri=http://<coordinator>:8080
http-server.http.port=8080
node-scheduler.include-coordinator=false
query.max-memory-per-node=8GB
query.max-memory=50GB

# if the coordinator is also a worker, it will have the following property instead
node-scheduler.include-coordinator=true

See Configuring the Presto Port for details on http port configuration.

Example

If you want to change the jvm configuration on the coordinator and the node.environment property from node.properties on all nodes, add the following jvm.config to ~/.prestoadmin/coordinator

-server
-Xmx16G
-XX:-UseBiasedLocking
-XX:+UseG1GC
-XX:G1HeapRegionSize=32M
-XX:+ExplicitGCInvokesConcurrent
-XX:+HeapDumpOnOutOfMemoryError
-XX:+UseGCOverheadLimit
-XX:OnOutOfMemoryError=kill -9 %p
-XX:ReservedCodeCacheSize=512M

Further, add the following node.properties to ~/.prestoadmin/coordinator and ~/.prestoadmin/workers:

node.environment=test
node.data-dir=/var/lib/presto/data
node.launcher-log-file=/var/log/presto/launcher.log
node.server-log-file=/var/log/presto/server.log
catalog.config-dir=/etc/presto/catalog
plugin.dir=/usr/lib/presto/lib/plugin

Then run:

./presto-admin configuration deploy

This will distribute to the coordinator a default config.properties, the new jvm.config and node.properties. The workers will receive the default config.properties and jvm.config, and the same node.properties as the coordinator.

If instead you just want to update the coordinator configuration, run:

./presto-admin configuration deploy coordinator

This will leave the workers configuration as it was, but update the coordinator’s configuration

configuration show

presto-admin configuration show [node|jvm|config|log]

This command prints the contents of the Presto configuration files deployed in the cluster. It takes an optional configuration name argument for the configuration files node.properties, jvm.config, config.properties and log.properties. For missing configuration files a warning will be printed except for log.properties file, since it is an optional configuration file in your Presto cluster.

If no argument is specified, then all four configurations will be printed.

Example

./presto-admin configuration show node

package install

presto-admin package install local_path [--nodeps]

This command copies any rpm from local_path to all the nodes in the cluster and installs it. Similar to server install the cluster topology is obtained from the file ~/.prestoadmin/config.json. If this file is missing, then the command prompts for user input to get the topology information.

This command takes an optional --nodeps flag which indicates if the rpm installed should ignore checking any package dependencies.

Warning

Using --nodeps can result in installing the rpm even with any missing dependencies, so you may end up with a broken rpm installation.

Example

./presto-admin package install /tmp/jdk-8u45-linux-x64.rpm

package uninstall

presto-admin package uninstall rpm_package_name [--nodeps]

This command uninstalls an rpm package from all the nodes in the cluster. Similar to server uninstall the cluster topology is obtained from the file ~/.prestoadmin/config.json. If this file is missing, then the command prompts for user input to get the topology information.

This command takes an optional --nodeps flag which indicates if the rpm installed should ignore checking any package dependencies.

Warning

Using --nodeps can result in uninstalling the rpm even when dependant packages are installed. It may end up with a broken rpm installation.

Example

./presto-admin package uninstall jdk

plugin add_jar

presto-admin plugin add_jar <local-path> <plugin-name> [<plugin-dir>]

This command deploys the jar at local-path to the plugin directory for plugin-name. By default /usr/lib/presto/lib/plugin is used as the top-level plugin directory. To deploy the jar to a different location, use the optional plugin-dir argument.

Example

./presto-admin plugin add_jar connector.jar my_connector
./presto-admin plugin add_jar connector.jar my_connector /my/plugin/dir

The first example will deploy connector.jar to /usr/lib/presto/lib/plugin/my_connector/connector.jar The second example will deploy it to /my/plugin/dir/my_connector/program.jar.

script run

presto-admin script run <local-path-to-script> [<remote-dir-to-put-script>]

This command can be used to run an arbitrary script on a cluster. It copies the script from its local location to the specified remote directory (defaults to /tmp), makes the file executable, and runs it.

Example

./presto-admin script run /my/local/script.sh
./presto-admin script run /my/local/script.sh /remote/dir

server install

presto-admin server install <rpm_specifier> [--rpm-source] [--nodeps]

This command takes in a parameter rpm_specifier. The parameter can be one of the following forms, listed in order of decreasing precedence: ‘latest’ - This downloads of the latest version of the presto rpm. url - This downloads the presto rpm found at the given url. version number - This downloads the presto rpm of the specified version. local path - This uses a previously downloaded rpm. The local path should be accessible by presto-admin. If rpm_specifier matches multiple forms, it is interpreted only as the form with highest precedence. For forms that require the rpm to be downloaded, if a local copy is found with a matching version to the rpm that would be downloaded, the local copy is used. Rpms downloaded using a version number or ‘latest’ come from Maven Central. This command fails if it cannot find or download the requested presto-server rpm.

After successfully finding the rpm, this command copies the presto-server rpm to all the nodes in the cluster, installs it, deploys the general presto configuration along with tpch connector configuration. The topology used to configure the nodes are obtained from ~/.prestoadmin/config.json. See Configuring Presto-Admin on how to configure your cluster using config.json. If this file is missing, then the command prompts for user input to get the topology information.

The general configurations for Presto’s coordinator and workers are taken from the directories ~/.prestoadmin/coordinator and ~/.prestoadmin/workers respectively. If these directories or any required configuration files are absent when you run server install, a default configuration will be deployed. See configuration deploy for details.

The catalog directory ~/.prestoadmin/catalog/ should contain the configuration files for any catalogs that you would like to connect to in your Presto cluster. The server install command will configure the cluster with all the catalogs in the directory. If the directory does not exist or is empty prior to server install, then by default the tpch connector is configured. See catalog add on how to add catalog configuration files after installation.

This command takes an optional --nodeps flag which indicates if the rpm installed should ignore checking any package dependencies.

Warning

Using --nodeps can result in installing the rpm even with any missing dependencies, so you may end up with a broken rpm installation.

Example

./presto-admin server install /tmp/presto.rpm
./presto-admin server install 0.148
./presto-admin server install http://search.maven.org/remotecontent?filepath=com/facebook/presto/presto-server-rpm/0.150/presto-server-rpm-0.150.rpm
./presto-admin server install latest

Standalone RPM Install

If you want to do a single node installation where coordinator and worker are co-located, you can just use:

rpm -i presto.rpm

This will deploy the necessary configurations for the presto-server to operate in single-node mode.

server restart

presto-admin server restart

This command first stops any Presto servers running and then starts them. A status check is performed on the entire cluster and is reported at the end.

Example

./presto-admin server restart

server start

presto-admin server start

This command starts the Presto servers on the cluster. A status check is performed on the entire cluster and is reported at the end.

Example

./presto-admin server start

server status

presto-admin server status

This command prints the status information of Presto in the cluster. This command will fail to report the correct status if the Presto installed is older than version 0.100. It will not print any status information if a given node is inaccessible.

The status output will have the following information:
  • server status
  • node uri
  • Presto version installed
  • node is active/inactive
  • catalogs deployed

Example

./presto-admin server status

server stop

presto-admin server stop

This command stops the Presto servers on the cluster.

Example

./presto-admin server stop

server uninstall

presto-admin server uninstall [--nodeps]

This command stops the Presto server if running on the cluster and uninstalls the Presto rpm. The uninstall command removes any presto related files deployed during server install but retains the Presto logs at /var/log/presto.

This command takes an optional --nodeps flag which indicates if the rpm uninstalled should ignore checking any package dependencies.

Example

./presto-admin server uninstall

server upgrade

presto-admin server upgrade path/to/new/package.rpm [local_config_dir] [--nodeps]

This command upgrades the Presto RPM on all of the nodes in the cluster to the RPM at path/to/new/package.rpm, preserving the existing configuration on the cluster. The existing cluster configuration is saved locally to local_config_dir (which defaults to a temporary folder if not specified). The path can either be absolute or relative to the current directory.

This command can also be used to downgrade the Presto installation, if the RPM at path/to/new/package.rpm is an earlier version than the Presto installed on the cluster.

Note that if the configuration files on the cluster differ from the presto-admin configuration files found in ~/.prestoadmin, the presto-admin configuration files are not updated.

This command takes an optional --nodeps flag which indicates if the rpm upgrade should ignore checking any package dependencies.

Warning

Using --nodeps can result in installing the rpm even with any missing dependencies, so you may end up with a broken rpm upgrade.

Example

./presto-admin server upgrade path/to/new/package.rpm /tmp/cluster-configuration
./presto-admin server upgrade /path/to/new/package.rpm /tmp/cluster-configuration

topology show

presto-admin topology show

This command shows the current topology configuration for the cluster (including the coordinators, workers, SSH port, and SSH username).

Example

./presto-admin topology show