Skip to content

Latest commit

 

History

History
181 lines (110 loc) · 5.63 KB

README.md

File metadata and controls

181 lines (110 loc) · 5.63 KB

Spark Cluster with Docker & docker-compose(2021 ver.)

General

A simple spark standalone cluster for your testing environment purposses. A docker-compose up away from you solution for your spark development environment.

The Docker compose will create the following containers:

container Exposed ports
spark-master 9090 7077
spark-worker-1 9091
spark-worker-2 9092
demo-database 5432

Installation

The following steps will make you run your spark cluster's containers.

Pre requisites

  • Docker installed

  • Docker compose installed

Build the image

docker build -t cluster-apache-spark:3.0.2 .

Run the docker-compose

The final step to create your test cluster will be to run the compose file:

docker-compose up -d

Validate your cluster

Just validate your cluster accesing the spark UI on each worker & master URL.

Spark Master

http://localhost:9090/

alt text

Spark Worker 1

http://localhost:9091/

alt text

Spark Worker 2

http://localhost:9092/

alt text

Resource Allocation

This cluster is shipped with three workers and one spark master, each of these has a particular set of resource allocation(basically RAM & cpu cores allocation).

  • The default CPU cores allocation for each spark worker is 1 core.

  • The default RAM for each spark-worker is 1024 MB.

  • The default RAM allocation for spark executors is 256mb.

  • The default RAM allocation for spark driver is 128mb

  • If you wish to modify this allocations just edit the env/spark-worker.sh file.

Binded Volumes

To make app running easier I've shipped two volume mounts described in the following chart:

Host Mount Container Mount Purposse
apps /opt/spark-apps Used to make available your app's jars on all workers & master
data /opt/spark-data Used to make available your app's data on all workers & master

This is basically a dummy DFS created from docker Volumes...(maybe not...)

Run Sample applications

NY Bus Stops Data [Pyspark]

This programs just loads archived data from MTA Bus Time and apply basic filters using spark sql, the result are persisted into a postgresql table.

The loaded table will contain the following structure:

latitude longitude time_received vehicle_id distance_along_trip inferred_direction_id inferred_phase inferred_route_id inferred_trip_id next_scheduled_stop_distance next_scheduled_stop_id report_hour report_date
40.668602 -73.986697 2014-08-01 04:00:01 469 4135.34710710144 1 IN_PROGRESS MTA NYCT_B63 MTA NYCT_JG_C4-Weekday-141500_B63_123 2.63183804205619 MTA_305423 2014-08-01 04:00:00 2014-08-01

To submit the app connect to one of the workers or the master and execute:

/opt/spark/bin/spark-submit --master spark://spark-master:7077 \
--jars /opt/spark-apps/postgresql-42.2.22.jar \
--driver-memory 1G \
--executor-memory 1G \
/opt/spark-apps/main.py

alt text

MTA Bus Analytics[Scala]

This program takes the archived data from MTA Bus Time and make some aggregations on it, the calculated results are persisted on postgresql tables.

Each persisted table correspond to a particullar aggregation:

Table Aggregation
day_summary A summary of vehicles reporting, stops visited, average speed and distance traveled(all vehicles)
speed_excesses Speed excesses calculated in a 5 minute window
average_speed Average speed by vehicle
distance_traveled Total Distance traveled by vehicle

To submit the app connect to one of the workers or the master and execute:

/opt/spark/bin/spark-submit --deploy-mode cluster \
--master spark://spark-master:7077 \
--total-executor-cores 1 \
--class mta.processing.MTAStatisticsApp \
--driver-memory 1G \
--executor-memory 1G \
--jars /opt/spark-apps/postgresql-42.2.22.jar \
--conf spark.driver.extraJavaOptions='-Dconfig-path=/opt/spark-apps/mta.conf' \
--conf spark.executor.extraJavaOptions='-Dconfig-path=/opt/spark-apps/mta.conf' \
/opt/spark-apps/mta-processing.jar

You will notice on the spark-ui a driver program and executor program running(In scala we can use deploy-mode cluster)

alt text

Summary

  • We compiled the necessary docker image to run spark master and worker containers.

  • We created a spark standalone cluster using 2 worker nodes and 1 master node using docker && docker-compose.

  • Copied the resources necessary to run demo applications.

  • We ran a distributed application at home(just need enough cpu cores and RAM to do so).

Why a standalone cluster?

  • This is intended to be used for test purposes, basically a way of running distributed spark apps on your laptop or desktop.

  • This will be useful to use CI/CD pipelines for your spark apps(A really difficult and hot topic)

Steps to connect and use a pyspark shell interactively

  • Follow the steps to run the docker-compose file. You can scale this down if needed to 1 worker.
docker-compose up --scale spark-worker=1
docker exec -it docker-spark-cluster_spark-worker_1 bash
apt update
apt install python3-pip
pip3 install pyspark
pyspark

What's left to do?

  • Right now to run applications in deploy-mode cluster is necessary to specify arbitrary driver port.

  • The spark submit entry in the start-spark.sh is unimplemented, the submit used in the demos can be triggered from any worker