SMA Execution Guide

The SMA-Checkpoints feature is an extensive workflow, so this section was created in order to be a walkthrough for feature usage.

PySpark Input

The SMA-Checkpoints feature requires a PySpark workload as its entry point, since it depends on detecting the use of PySpark DataFrames. This walkthrough will guide you through the feature using a single Python script, providing a straightforward example of how checkpoints are generated and utilized within a typical PySpark workflow.

Input workload

Sample.py file content

from pyspark.sql import SparkSession
from pyspark.sql import functions as F

spark = SparkSession.builder.appName("SparkFunctionsExample2").getOrCreate()

df1 = spark.createDataFrame([("Alice", "NY"), ("Bob", "LA")], ["name", "city"])
df2 = spark.createDataFrame([(10,), (20,)], ["number"])

df1_with_index = df1.withColumn("index", F.monotonically_increasing_id())
df2_with_index = df2.withColumn("index", F.monotonically_increasing_id())

df3 = df1_with_index.join(df2_with_index, on="index").drop("index")
df3.show()

Migrating Workload

Feature Enabled

If the SMA-Checkpoints feature is enabled, a checkpoints.json file will be generated. If the feature is disabled, this file will not be created in either the input or output folders. Regardless of whether the feature is enabled, the following inventory files will always be generated: DataFramesInventory.csv and CheckpointsInventory.csv. These files provide metadata essential for analysis and debugging.

Conversion Process

To create a convert your own project please follow up the following guide: SMA User Guide.

SMA-Checkpoints Feature Settings

As part of the conversion process you can customize your conversion settings, take a look on the SMA-Checkpoints feature settings.

Note: This user guide used the default conversion settings.

Conversion Results

Once the migration process is complete, the SMA-Checkpoints feature should have created two new inventory files and added a checkpoints.json file to both the input and output folders.

Take a look on SMA-Checkpoints inventories to review the related inventories.

Input Folder

checkpoints.json file content

{
  "createdBy": "Snowpark Migration Accelerator",
  "comment": "This file was automatically generated by the SMA tool as checkpoints collection was enabled in the tool settings. This file may also be modified or deleted during SMA execution.",
  "type": "Collection",
  "pipelines": [
    {
      "entryPoint": "sample.py",
      "checkpoints": [
        {
          "name": "sample$BBVOC7$df1$1",
          "file": "sample.py",
          "df": "df1",
          "location": 1,
          "enabled": true,
          "mode": 1,
          "sample": "1.0"
        },
        {
          "name": "sample$BBVOC7$df2$1",
          "file": "sample.py",
          "df": "df2",
          "location": 1,
          "enabled": true,
          "mode": 1,
          "sample": "1.0"
        },
        {
          "name": "sample$BBVOC7$df3$1",
          "file": "sample.py",
          "df": "df3",
          "location": 1,
          "enabled": true,
          "mode": 1,
          "sample": "1.0"
        }
      ]
    }
  ]
}

Output Folder

checkpoints.json file content

{
  "createdBy": "Snowpark Migration Accelerator",
  "comment": "This file was automatically generated by the SMA tool as checkpoints collection was enabled in the tool settings. This file may also be modified or deleted during SMA execution.",
  "type": "Validation",
  "pipelines": [
    {
      "entryPoint": "sample.py",
      "checkpoints": [
        {
          "name": "sample$BBVOC7$df1$1",
          "file": "sample.py",
          "df": "df1",
          "location": 1,
          "enabled": true,
          "mode": 1,
          "sample": "1.0"
        },
        {
          "name": "sample$BBVOC7$df2$1",
          "file": "sample.py",
          "df": "df2",
          "location": 1,
          "enabled": true,
          "mode": 1,
          "sample": "1.0"
        },
        {
          "name": "sample$BBVOC7$df3$1",
          "file": "sample.py",
          "df": "df3",
          "location": 1,
          "enabled": true,
          "mode": 1,
          "sample": "1.0"
        }
      ]
    }
  ]
}

Once the SMA execution flow is complete and both the input and output folders contain their respective checkpoints.json files, you are ready to begin the Snowpark-Checkpoints execution process.