Pulsar SQL configuration and deployment

You can configure Presto Pulsar connector and deploy a cluster with the following instruction.

Configure Presto Pulsar Connector

You can configure Presto Pulsar Connector in the ${project.root}/conf/presto/catalog/pulsar.properties properties file. The configuration for the connector and the default values are as follows.

# name of the connector to be displayed in the catalog
connector.name=pulsar

# the url of Pulsar broker service
pulsar.broker-service-url=http://localhost:8080

# URI of Zookeeper cluster
pulsar.zookeeper-uri=localhost:2181

# minimum number of entries to read at a single time
pulsar.entry-read-batch-size=100

# default number of splits to use per query
pulsar.target-num-splits=4

You can connect Presto to a Pulsar cluster with multiple hosts. To configure multiple hosts for brokers, add multiple URLs to pulsar.broker-service-url. To configure multiple hosts for ZooKeeper, add multiple URIs to pulsar.zookeeper-uri. The following is an example.

pulsar.broker-service-url=http://localhost:8080,localhost:8081,localhost:8082
pulsar.zookeeper-uri=localhost1,localhost2:2181

Note: by default, Pulsar SQL does not get the last message in a topic. It is by design and controlled by settings. By default, BookKeeper LAC only advances when subsequent entries are added. If there is no subsequent entry added, the last written entry is not visible to readers until the ledger is closed. This is not a problem for Pulsar which uses managed ledger, but Pulsar SQL directly reads from BookKeeper ledger.

If you want to get the last message in a topic, set the following configurations:

For the broker configuration, set bookkeeperExplicitLacIntervalInMills > 0 in broker.conf or standalone.conf.
For the Presto configuration, set pulsar.bookkeeper-explicit-interval > 0 and pulsar.bookkeeper-use-v2-protocol=false.

However, using BookKeeper V3 protocol introduces additional GC overhead to BK as it uses Protobuf.

Query data from existing Presto clusters

If you already have a Presto cluster, you can copy the Presto Pulsar connector plugin to your existing cluster. Download the archived plugin package with the following command.

$ wget https://archive.apache.org/dist/pulsar/pulsar-2.7.1/apache-pulsar-2.7.1-bin.tar.gz

Deploy a new cluster

Since Pulsar SQL is powered by Presto, the configuration for deployment is the same for the Pulsar SQL worker.

note

For how to set up a standalone single node environment, refer to Query data.

You can use the same CLI args as the Presto launcher.

$ ./bin/pulsar sql-worker --help
Usage: launcher [options] command

Commands: run, start, stop, restart, kill, status

Options:
  -h, --help            show this help message and exit
  -v, --verbose         Run verbosely
  --etc-dir=DIR         Defaults to INSTALL_PATH/etc
  --launcher-config=FILE
                        Defaults to INSTALL_PATH/bin/launcher.properties
  --node-config=FILE    Defaults to ETC_DIR/node.properties
  --jvm-config=FILE     Defaults to ETC_DIR/jvm.config
  --config=FILE         Defaults to ETC_DIR/config.properties
  --log-levels-file=FILE
                        Defaults to ETC_DIR/log.properties
  --data-dir=DIR        Defaults to INSTALL_PATH
  --pid-file=FILE       Defaults to DATA_DIR/var/run/launcher.pid
  --launcher-log-file=FILE
                        Defaults to DATA_DIR/var/log/launcher.log (only in
                        daemon mode)
  --server-log-file=FILE
                        Defaults to DATA_DIR/var/log/server.log (only in
                        daemon mode)
  -D NAME=VALUE         Set a Java system property

The default configuration for the cluster is located in ${project.root}/conf/presto. You can customize your deployment by modifying the default configuration.

You can set the worker to read from a different configuration directory, or set a different directory to write data.

$ ./bin/pulsar sql-worker run --etc-dir /tmp/incubator-pulsar/conf/presto --data-dir /tmp/presto-1

You can start the worker as daemon process.

$ ./bin/pulsar sql-worker start

Deploy a cluster on multiple nodes

You can deploy a Pulsar SQL cluster or Presto cluster on multiple nodes. The following example shows how to deploy a cluster on three-node cluster.

Copy the Pulsar binary distribution to three nodes.

The first node runs as Presto coordinator. The minimal configuration requirement in the ${project.root}/conf/presto/config.properties file is as follows.

coordinator=true
node-scheduler.include-coordinator=true
http-server.http.port=8080
query.max-memory=50GB
query.max-memory-per-node=1GB
discovery-server.enabled=true
discovery.uri=<coordinator-url>

The other two nodes serve as worker nodes, you can use the following configuration for worker nodes.

coordinator=false
http-server.http.port=8080
query.max-memory=50GB
query.max-memory-per-node=1GB
discovery.uri=<coordinator-url>

Modify pulsar.broker-service-url and pulsar.zookeeper-uri configuration in the ${project.root}/conf/presto/catalog/pulsar.properties file accordingly for the three nodes.
Start the coordinator node.

$ ./bin/pulsar sql-worker run

Start worker nodes.

$ ./bin/pulsar sql-worker run

Start the SQL CLI and check the status of your cluster.

$ ./bin/pulsar sql --server <coordinate_url>

Check the status of your nodes.

presto> SELECT * FROM system.runtime.nodes;
 node_id |        http_uri         | node_version | coordinator | state
---------+-------------------------+--------------+-------------+--------
 1       | http://192.168.2.1:8081 | testversion  | true        | active
 3       | http://192.168.2.2:8081 | testversion  | false       | active
 2       | http://192.168.2.3:8081 | testversion  | false       | active

For more information about deployment in Presto, refer to Presto deployment.

note

The broker does not advance LAC, so when Pulsar SQL bypass broker to query data, it can only read entries up to the LAC that all the bookies learned. You can enable periodically write LAC on the broker by setting "bookkeeperExplicitLacIntervalInMills" in the broker.conf.