Configuring Presto

Presto configuration parameters can be modified to tweak performance or add/remove features. While Presto is designed to work well out-of-the-box, you still may need to make some changes.

Memory configuration

It is often necessary to change the default memory configuration based on your cluster’s capacity. The default max memory for each Presto server is 16 GB, but if you have a lot of memory (say, 120GB/node), you may want to allocate more memory to Presto for better performance.

In order to update the max memory value to 60 GB per node:

  1. Change the line in ~/.prestoadmin/coordinator/jvm.config and ~/.prestoadmin/workers/jvm.config that says -Xmx16G to -Xmx60G.

  2. Change the following lines in ~/.prestoadmin/coordinator/config.properties and ~/.prestoadmin/workers/config.properties:

    query.max-memory-per-node=8GB
    query.max-memory=50GB
    

    to

    query.max-memory-per-node=30GB
    query.max-memory=<30GB * number of nodes>
    

    We recommend setting query.max-memory-per-node to half of the JVM config max memory, though if your workload is highly concurrent, you may want to use a lower value for query.max-memory-per-node. If you have large data skew, query.max-memory-per-node should. By default in Presto 148t and higher, query.max-memory-per-node is 10% of the Xmx value specified in jvm.config.

  3. Run the following command to deploy the configuration change to the cluster:

    ./presto-admin configuration deploy
    
  4. Restart the Presto servers so that the changes get picked up:

    ./presto-admin server restart
    

    If you are running Presto in a test environment that has less than 16 GB of memory available, you will need to follow similar procedures to set the memory configurations lower.

Log file location configurations

For most production environments, it will be necessary to change the log locations. In order to update these:

  1. Stop the Presto server.

    ./presto-admin server stop
    
  2. Presto stores logs and other data in node.data-dir, node.launcher-log-file, and node.server-log-file. It is very important that these locations have enough space for the logs on the filesystem on each node where Presto is running. The default location for node.data-dir is /var/lib/presto/data, the default location for node.launcher-log-file is /var/log/presto/launcher.log, and the default location for node.server-log-file is /var/log/presto/server.log. Assuming the chosen locations are /data1/presto and /data2/presto for the data directory and server logs respectively, the properties in ~/.prestoadmin/coordinator/node.properties and ~/.prestoadmin/workers/node.properties will be as follows:

    node.data-dir=/data1/presto/data
    node.launcher-log-file=/data2/presto/launcher.log
    node.server-log-file=/data2/presto/server.log
    
  3. The log directory(ies) (in the above example, /data1/presto and /data2/presto; the data directory for node.data-dir is created by Presto) need to exist on all nodes and be owned by the presto user. The command presto-admin run_script can be used to perform these actions on all of the nodes. First, create a script in the same directory as presto-admin, called script.sh:

    #!/bin/bash
    mkdir -p /data1/presto
    mkdir -p /data2/presto
    chown presto:presto /data1/presto
    chown presto:presto /data2/presto
    

    Then, run the following command:

    ./presto-admin run_script script.sh
    
  4. Run the following command to deploy the log configuration change to the cluster:

    ./presto-admin configuration deploy
    
  5. Restart the Presto servers so that the changes get picked up:

    ./presto-admin server restart
    

For detailed documentation on configuration deploy, see configuration deploy. For more configuration parameters, see the Presto documentation.