Link Search Menu Expand Document

Using lakeFS with Spark

Apache Spark is a unified analytics engine for big data processing, with built-in modules for streaming, SQL, machine learning and graph processing.

Table of contents

  1. Configuration
  2. Accessing Objects
  3. Creating Objects

Accessing data in lakeFS from Spark is the same as accessing S3 data from Spark. The only changes we need to consider are:

  1. Setting the configurations to access lakeFS
  2. Accessing Objects with the lakeFS path convention


In order to configure Spark to work with lakeFS we will set the lakeFS credentials in the corresponding S3 credential fields.

lakeFS endpoint: fs.s3a.endpoint

lakeFS access key: fs.s3a.access.key

lakeFS secret key: fs.s3a.secret.key

Note In the following examples we set AWS credentials at runtime, for clarity. In production, these properties should be set using one of Hadoop’s standard ways of Authenticating with S3.

For example if we would like to Specify the credentials at run time

spark.sparkContext.hadoopConfiguration.set("fs.s3a.access.key", "AKIAIOSFODNN7EXAMPLE")
spark.sparkContext.hadoopConfiguration.set("fs.s3a.secret.key", "wJalrXUtnFEMI/K7MDENG/bPxRfiCYEXAMPLEKEY")
spark.sparkContext.hadoopConfiguration.set("fs.s3a.endpoint", "")

Accessing Objects

In order for us to access objects in lakeFS we will need to use the lakeFS path conventions: s3a://[REPOSITORY]/[BRANCH]/PATH/TO/OBJECT

For example: Lets assume we want to read a parquet file:

from repository: example-repo branch: master in the path: example-path

val repo = "example-repo"
val branch = "master"
val dataPath = s"s3a://${repo}/${branch}/example-path/example-file.parquet"
val basics =

Creating Objects

If we would like to create new parquet files partitioned by column example-column