Presto Tutorial 103: PrestoDB cluster on GCP

Introduction

This tutorial is Part III of our Getting started with PrestoDB series. As a reminder, Prestodb is an open source distributed SQL query engine. In tutorial 102 we covered how to run a three node prestodb cluster on a laptop. In this tutorial, we’ll show you how to run a prestodb cluster in a GCP environment using VM instances and GKE containers.

Environment

This guide was developed on GCP VM instances and GKE containers.

Presto on GCP with VMs

Implementation steps for prestodb on vm instances

Step1: Create a GCP VM instance using the CREATE INSTANCE tab, name it as presto-coordinator. Next, create three more VM instances as presto-worker1, presto-worker2 and presto-worker3 respectively.

Step 2: By default GCP blocks all network ports, so prestodb will need ports 8080-8083 enabled. Use the firewalls rule tab and enable them.

Step 3: 

Install JAVA and python.

Step 4:

Download the Presto server tarball, presto-server-0.253.1.tar.gz and unpack it. The tarball will contain a single top-level directory, presto-server-0.253.1 which we will call the installation directory.

Run the commands below to install the official tarballs for presto-server and presto-cli from prestodb.io

user@presto-coordinator-1:~$ curl -O https://repo1.maven.org/maven2/com/facebook/presto/presto-server/0.235.1/presto-server-0.235.1.tar.gz
 % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                Dload  Upload   Total   Spent    Left  Speed
100  721M  100  721M    0     0   245M      0  0:00:02  0:00:02 --:--:--  245M
user@presto-coordinator-1:~$ curl -O https://repo1.maven.org/maven2/com/facebook/presto/presto-cli/0.235.1/presto-cli-0.235.1-executable.jar
 % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                Dload  Upload   Total   Spent    Left  Speed
100 12.7M  100 12.7M    0     0  15.2M      0 --:--:-- --:--:-- --:--:-- 15.1M
user@presto-coordinator-1:~$

Step 5:

Use gunzip and tar to unzip and untar the presto-server

Step 6: (optional)

Rename the directory without version number

Step 7:  

Create etc, etc/catalog and data directories

Step 8:

Define etc/node.config, etc/config.properties, etc/jvm.config and etc/catalog/jmx.properties files as below for presto co-ordinator server.  

user@presto-coordinator-1:~/presto-server$ cat etc/node.properties
node.environment=production
node.id=ffffffff-ffff-ffff-ffff-ffffffffffff
node.data-dir=/home/user/presto-server/data

user@presto-coordinator-1:~/presto-server$ cat etc/config.properties
coordinator=true
node-scheduler.include-coordinator=false
http-server.http.port=8080
query.max-memory=50GB
query.max-memory-per-node=1GB
query.max-total-memory-per-node=2GB
discovery-server.enabled=true
discovery.uri=http://localhost:8080

user@presto-coordinator-1:~/presto-server$ cat etc/jvm.config
-server-Xmx16G
-XX:+UseG1GC
-XX:G1HeapRegionSize=32M
-XX:+UseGCOverheadLimit
-XX:+ExplicitGCInvokesConcurrent
-XX:+HeapDumpOnOutOfMemoryError
-XX:+ExitOnOutOfMemoryError
-Djdk.attach.allowAttachSelf=true

user@presto-coordinator-1:~/presto-server$ cat etc/log.properties
com.facebook.presto=INFO

user@presto-coordinator-1:~/presto-server$ cat etc/catalog/jmx.properties
connector.name=jmx

Step: 9 

Check the cluster UI status. It should  show the Active worker count at 0 since we enabled only the coordinator.

Step 10: 

Repeat steps 1 to 8 on the remaining 3 vm instances which will act as worker nodes.

On the configuration step for worker nodes, set coordinator to false and http-server.http.port to 8081, 8082 and 8083 for worker1, worker2 and worker3 respectively.

Also make sure node.id and http-server.http.port are different for each worker node.

user@presto-worker1:~/presto-server$ cat etc/node.properties
node.environment=production
node.id=ffffffff-ffff-ffff-ffff-ffffffffffffd
node.data-dir=/home/user/presto-server/data
user@presto-worker1:~/presto-server$ cat etc/config.properties
coordinator=false
http-server.http.port=8083
query.max-memory=50GB
query.max-memory-per-node=1GB
query.max-total-memory-per-node=2GB
discovery.uri=http://presto-coordinator-1:8080

user@presto-worker1:~/presto-server$ cat etc/jvm.config
-server-Xmx16G
-XX:+UseG1GC
-XX:G1HeapRegionSize=32M
-XX:+UseGCOverheadLimit
-XX:+ExplicitGCInvokesConcurrent
-XX:+HeapDumpOnOutOfMemoryError
-XX:+ExitOnOutOfMemoryError
-Djdk.attach.allowAttachSelf=true

user@presto-worker1:~/presto-server$ cat etc/log.properties
com.facebook.presto=INFO

user@presto-worker1:~/presto-server$ cat etc/catalog/jmx.properties
connector.name=jmx

Step 11: 

Check cluster status, it should reflect the three worker nodes as part of the prestodb cluster.

Step 12:

Verify the prestodb environment by running the prestodb CLI with simple JMX query

Implementation steps for Prestodb on GKE containers

Step 1:

Go to the Google cloud Console and activate the cloud console window

Step 2:

Create an artifacts repository using the below command and replace REGION with the valid region you would prefer to create the repository.

Step 3:

Create the container cluster by using the gcloud command: 

Step 4:

After container cluster creation, run the following command to see the cluster’s three nodes

Step 5:

Pull the prestodb docker image 

Step 6:

Deploy ahanaio/prestodb-sandbox locally on the shell and create an image named as coordinator which will later be deployed on the container clusters.

Step 7:

Edit the etc/config.parameters file inside the container and set the node-scheduler.include-coordinator property to false. Now restart the coordinator.

Step 8:

Now do a docker commit, create a tag called coordinator based on imageid, this will create a new local image called coordinator.

Step 9:

Create tag with artifacts path and copy it over to artifacts location

Step 10:

Deploy the coordinator into the cloud container using the below kubectl commands.

Step 11:

Copy the external IP on a browser and check the status

Step 12:

Now to deploy worker1 into the GKE container, again start a local instance named worker1 using the docker run command.

Step 13:

Edit worker1 config.properties inside the worker1 container to set coordinator to false and http-server.http.port to 8081. Also the discovery.uri should point to the coordinator container running inside the GKE container.

Step 14:

Stop the local worker1 container, commit the worker1 as image and tag it as worker1 image

Step 15:

Push the worker1 image into google artifacts location

Step 16:

Deploy and expose the worker1 from the artifacts location into the google cloud container using this kubectl command.

Step 17:

Check presto UI for successful deployment of worker1

Step 18:

Repeat steps 12 to steps 17 to deploy worker2 inside GKE container:

• deploy ahana local instance using docker and name it as worker2, 
• then edit the etc/config.properties inside the worker2 container to set coordinator to false, port to 8082 and discover.uri to the coordinator container name.
• shut the instance then commit that instance and create docker image as worker2 
• push that worker2 image to google artifacts location 
• use kubectl commands to deploy and expose the worker2 instance inside a google container. Now check the prestodb UI for the second worker being active.
• Check prestodb UI for successful deployment of worker2

Steps 19:

Repeat steps 12 to steps 18 to provision worker3 inside the google cloud container

Step 20:

Verify the prestodb environment by running the prestodb CLI with simple JMX query

Summary

In this tutorial you learned how to  provision and run prestodb inside Google VM instances and on GKE containers. Now you should be able to validate the functional aspects of prestodb. 

If you want to run production Presto workloads at scale and performance, check out https://www.ahana.io which provides a managed service for Presto.