An end-to-end Azure Data Factory (ADF) project demonstrating real-world data engineering patterns — REST API ingestion, on-premises file ingestion, incremental loading, watermark management, orchestration, alerting, audit logging, and REST API pagination.

I've documented the incremental loading concepts and implementation behind this project in a 4-part Medium series. Click a button below to jump straight to the article.

💡 For the incremental pipelines in this repository, start with Part 1 and follow the series sequentially.

This project contains 7 Azure Data Factory pipelines, each covering a common real-world data engineering scenario.

pl_API_Ingestion

Pulls DimAirport.json from GitHub Raw content into the ADLS Gen2 Bronze layer.

GitHub Raw → Copy API Data → ADLS Gen2 Bronze

• Source: ds_APISource → ls_githubAPI_source → HttpServer / Anonymous → DimAirport.json
• Sink: ds_github_sink → ls_bronze → ADLS Gen2 → bronze/github/DimAirport.json
• Key Settings: JsonSource with HttpReadSettings, HTTP GET, JsonSink with AzureBlobFSWriteSettings, retry = 2, retryIntervalInSeconds = 30, enableStaging = false

pl_OnPremFiles_To_AzureBlobBronze

Migrates 3 CSV files in parallel from an on-premises Windows file server to ADLS Gen2 Bronze: DimAirline.csv, DimFlight.csv, DimPassenger.csv.

On-Prem Windows File Server → ForEach (Parallel) → Copy Activity → ADLS Gen2 Bronze

• Pipeline Parameter: file_name — e.g. [{"fileName": "DimAirline.csv"}, {"fileName": "DimFlight.csv"}, {"fileName": "DimPassenger.csv"}]
• Key Expressions: @pipeline().parameters.file_name, @item().fileName, @dataset().p_fileName_source, @dataset().p_fileName_sink
• Source: ds_OnPremSource → ls_onPrem_Files → FileServer / IROnPremise → E:\OnPremiseFiles\{fileName}
• Sink: ds_onPremSink → ls_bronze → ADLS Gen2 → bronze/OnPremDataFiles/{fileName}

pl_SQLtoAzure_Incremental_Json

Performs an incremental load from Azure SQL FactBookings to ADLS Gen2 in Parquet format, using a JSON file as the watermark store.

LastLoad Lookup ─┐
                  ├─→ Copy SQL Data → Parquet
LatestLoad Lookup ┘
                        ↓
                 Update Watermark

• Last Load: reads lastload.json from Azure Blob to get the last processed booking_date
• Latest Load: SELECT MAX(booking_date) AS LatestLoad FROM dbo.FactBookings;
• Incremental Query:

  SELECT * FROM dbo.FactBookings
  WHERE booking_date > '@{activity('LastLoad').output.firstRow.lastload}'
  AND booking_date <= '@{activity('LatestLoad').output.firstRow.LatestLoad}';

• Sink: Azure SQL → Copy Activity → ADLS Gen2 → Bronze/sql/ → Parquet + Snappy compression
• Watermark Update: the latest watermark is written back to lastload.json

📖 Detailed explanation in Part 1 of the Medium series above.

pl_dailydata_load_with_alerts

Orchestrator pipeline that calls the incremental pipeline and reports execution status to an Azure Logic App.

DailyDataLoad → Execute Pipeline → pl_SQLtoAzure_Incremental_Json → Call Logic App

• Execute Pipeline: calls pl_SQLtoAzure_Incremental_Json with waitOnCompletion = true
• Logic App Trigger: fires on both ✅ Success and ❌ Failure
• Payload:

  {
    "pipeline_name": "@{pipeline().Pipeline}",
    "run_id": "@{pipeline().RunId}",
    "status": "@{activity('DailyDataLoad').Status}",
    "error": "@{if(equals(activity('DailyDataLoad').Status,'Failed'), string(activity('DailyDataLoad').error), 'No Error')}"
  }

pl_onPremSQl_To_AzureSQL_Incremental_WatermarkTable

Multi-table incremental loading from On-Premises SQL Server to Azure SQL using a centralized watermark table, for dbo.iplteams and dbo.salesitems, both keyed on a last_updated watermark column.

ForEach Table
     ↓
Old Watermark ─┐
               ├→ If Condition ─→ TRUE → Copy → Update Watermark
New Watermark ─┘                → FALSE → Skip

• Old Watermark: SELECT WatermarkValue FROM watermarktable WHERE TableName = '@{item().TableName}';
• New Watermark: SELECT MAX(last_updated) AS NewWatermarkValue FROM @{item().TableName};
• If Condition: @greater(activity('New_LookUp_Col').output.firstRow.NewWatermarkValue, activity('Old_LookUp_Col').output.firstRow.WatermarkValue)
• Dynamic SQL built from table schema, table name, watermark column, old and new watermark values
• Sink: AzureSqlSink, writeBehavior = upsert, useTempDB = true, keys = [id]
• Watermark Update: stored procedure [dbo].[usp_write_watermark] with parameters last_updated, tableName

📖 Detailed explanation in the Medium series above.

pl_onPremSQl_To_AzureSQL_Incremental_WatermarkTable_ScriptActivity

An enhanced version of Pipeline 5 adding execution timing, a Script Activity for watermark updates, success/failure audit logging, rows-copied tracking, and error tracking.

Set Start Time → Old Watermark Lookup → New Watermark Lookup → If Condition
                                                                     ↓
                                                          Copy → Success/Failed → Audit

• Set Start Time: PipelineStartTime = @utcNow()
• Script Activity — Update Watermark:

  UPDATE dbo.watermarktable
  SET WatermarkValue = @last_updated
  WHERE TableName = @tableName;

• Success/Failure Audit writes to dbo.PipelineAudit: PipelineName, TableName, SourceName, TargetName, WatermarkOld, WatermarkNew, RowsCopied, Status, ErrorMessage, StartTime, EndTime
• Key Expressions: @pipeline().Pipeline, @item().TableName, @activity('Copy_FromOnPremSQL_ToAzureSQL').output.rowsCopied, @activity('Copy_FromOnPremSQL_ToAzureSQL').error.message, @variables('PipelineStartTime'), @utcNow()
• Evolution from Pipeline 5: the old SP_Update_WaterMark activity is now inactive — watermark updates are handled by the Script Activity instead

📖 Detailed explanation in the Medium series above.

pl_paginationExample

Ingests Pokémon data from the PokéAPI across multiple pages using ADF's native RANGE pagination.

PokéAPI → Web Activity → Get Total Count → Copy Activity → RANGE Pagination → ADLS Gen2 Bronze

• Initial Request: https://pokeapi.co/api/v2/pokemon?limit=20&offset=0
• Pagination Rule: QueryParameters.offset = RANGE:0:@{activity('Get API Data').output.count}:20 → generates offset=0, 20, 40, ... 1300
• Sink: ds_API_Target_JSON → ADLS Gen2 → bronze/PaginationExample/

REST API ingestion · ADLS Gen2 · On-Premises File Server · Self-hosted Integration Runtime · ForEach activity · Parallel processing · Pipeline parameters · Dataset parameters · Dynamic expressions · Incremental loading · JSON watermark · Watermark table · Azure SQL Upsert · Execute Pipeline · Web Activity · Logic App integration · Script Activity · Audit logging · Error handling · REST API pagination · RANGE pagination

I built multiple Azure Data Factory pipelines covering real-world data engineering scenarios — REST API ingestion, parallel on-premises file migration, watermark-based incremental loading, multi-table processing using ForEach, Azure SQL upserts, Logic App alerts, Script Activity-based watermark management, audit logging, and REST API pagination.

github.com/Prabhas92/ADFProject

Chethan Prabhas

Azure Data Factory • SQL • Azure • Python • Data Engineering