spark-analyzer 0.1.23

A tool for analyzing Spark History Server data and identifying optimization opportunities

Spark Analyzer

Analyze your Apache Spark applications for performance bottlenecks, resource inefficiencies, and optimization opportunities. Get actionable insights from your Spark History Server data with detailed stage-level analysis and recommendations.

Quick Start - No Registration: Run spark-analyzer analyze --save-local for immediate analysis. Enhanced Analysis (optional): Obtain your ID for cost estimation and deeper analysis of your data.

Features

• Workflow Classification: Automatically categorizes Extract, Transform, and Load operations in Spark jobs
• Stage-Level Analysis: Detailed performance metrics and resource utilization analysis
• Storage Format Detection: Identifies Parquet, Delta, Hudi, and Iceberg usage patterns
• Privacy Controls: Configurable data hashing for masking sensitive information
• Local & Browser Modes: Support for direct access and browser-based Spark History Server interfaces
• Offline Analysis: Run completely offline with --save-local flag for local data processing and reporting. No data leaves your environment.
• Report Generation: Excel output with detailed optimization recommendations

What You'll Get

Real analysis showing 34.8% compute waste and 45-65% performance improvement opportunities

Ready to analyze your Spark applications?

Quick Start

# Install
pip install spark-analyzer

# Configure connection to your Spark History Server
spark-analyzer configure

# Local analysis (no registration required)
spark-analyzer analyze --save-local

# Generate Excel report
spark-analyzer report --input analysis_data.json --output my_report.xlsx

# Enhanced analysis with cost insights (if you have an ID)
spark-analyzer analyze

Installation

# Create and activate a virtual environment (recommended)
python3 -m venv myenv
source myenv/bin/activate  # On Windows: myenv\Scripts\activate

# Install the package
pip install spark-analyzer

Note: If you encounter "externally-managed-environment" errors, use a virtual environment as shown above.

Commands

The tool uses a subcommand structure for organization:

# Main commands
spark-analyzer                   # Default: runs analyze command
spark-analyzer analyze           # Explicit analyze command
spark-analyzer configure         # Configuration wizard
spark-analyzer report            # Generate acceleration reports
spark-analyzer readme            # Show the README documentation

# Get help
spark-analyzer --help            # Show general help
spark-analyzer analyze --help    # Show analyze command help
spark-analyzer configure --help  # Show configure command help
spark-analyzer report --help     # Show report command help

Usage

Basic Analysis

# Run with saved configuration
spark-analyzer

# Run with explicit analyze command
spark-analyzer analyze

# Run in local mode
spark-analyzer analyze --local

# Run in browser mode
spark-analyzer analyze --browser

# Run with local data saving (no upload)
spark-analyzer analyze --save-local

# Run with privacy protection
spark-analyzer analyze --opt-out name,description,details

Configuration

# Basic configuration
spark-analyzer configure

First-time users: The tool automatically runs the configuration wizard to set up connection mode and Spark History Server URL. Cost Estimator ID is optional for enhanced features.

Subsequent runs: Uses saved configuration and runs immediately.

Configuration file: Settings are saved to ~/.spark_analyzer/config.ini and can be manually edited if needed.

Runtime options (not saved in configuration):

• --opt-out: Privacy protection settings
• --save-local: Local data saving without upload

Documentation: Use spark-analyzer readme to view the full documentation in your terminal.

Report Generation

# Generate Excel report (no data leaves your environment)
spark-analyzer report --input your_data.json --output report.xlsx

# Batch processing multiple files
spark-analyzer report --input "*.json" --output-dir ./reports/

# Get help
spark-analyzer report --help

The Excel report provides comprehensive Spark performance analysis. Optional: Send your report to [e-mail] for advanced optimization recommendations or technical consultation.

Connection Modes

The tool supports two connection modes:

Local Mode

Use when you have direct access to the Spark History Server (port forwarding or SSH tunnel).

1. Configure:

   spark-analyzer configure
   

2. Manual Configuration (optional): Edit ~/.spark_analyzer/config.ini:

   [server]
   # Standard installation
   base_url = http://localhost:18080
   
   # Port forwarding
   # base_url = http://localhost:8080/onehouse-spark-code/history-server
   

3. Run:

   spark-analyzer analyze --save-local  # Local analysis
   # or
   spark-analyzer analyze               # Enhanced analysis (prompts for ID if needed)
   

Browser Mode

Use when accessing Spark History Server through a browser (e.g., EMR interface).

1. Configure:

   spark-analyzer configure
   

2. Runtime Configuration: URLs and cookies are collected at runtime for security.

3. Get Browser Cookies:

   • Open Spark History Server in browser
   • Open developer tools (F12)
   • Go to Network tab and click any request
   • Find "Cookie" header in Request Headers
   • Copy the entire cookie string

4. Run:

   spark-analyzer analyze --save-local  # Local analysis
   # or
   spark-analyzer analyze               # Enhanced analysis (prompts for ID if needed)
   

Supported Platforms

The tool supports various Spark environments including EMR, Databricks, and other Spark History Server implementations. The configuration wizard will guide you through the specific setup steps for your environment.

Data Collection

Collected Information

Application-Level Data

• Application ID, name, start/end times
• Executor count and core allocation
• CPU and memory usage metrics

Stage-Level Data

• Stage ID, attempt information, names, descriptions
• Task counts and executor assignments
• Performance metrics (duration, CPU time)
• Execution plans and stack traces

Derived Metrics

• Workflow type classification (Extract, Transform, Load)
• Storage format detection
• Resource utilization analysis
• Performance bottleneck identification

Privacy Controls

Use --opt-out to hash sensitive fields:

# Hash all potentially sensitive fields
spark-analyzer analyze --opt-out name,description,details

# Hash only stage names
spark-analyzer analyze --opt-out name

# Hash only stage descriptions
spark-analyzer analyze --opt-out description

Available Options:

• name: Stage names → name_hash_[numeric_hash]
• description: Stage descriptions → description_hash_[numeric_hash]
• details: Execution plans → details_hash_[numeric_hash]

Example:

• Original: "stage_name": "show at Console.scala:14"
• Hashed: "stage_name": "name_hash_1002607082777652347"

Data Usage

• Purpose: Performance analysis and optimization recommendations
• Processing: Remote analysis data is processed by Onehouse's cloud infrastructure to generate personalized reports. With --save-local mode, all processing happens locally - no data is transmitted.
• Security: Data transmitted over HTTPS by configuring cost estimator ID is stored in Onehouse's secure cloud environment
• Retention: Analysis data is deleted once the report is emailed to you

Enterprise Considerations

Data Collection

• Uses Python's built-in hash() function for consistent hashing
• Tool does not collect any PII. Only stated data from Spark is processed/collected.
• Stage content may contain file paths, table names, or query fragments. Turn on privacy options if they are sensitive.
• Anonymous usage data collected via Scarf for tool improvement and analytics

Report Output

Excel Report Structure

• Application Level: Overall metrics and acceleration summary
• Workflow Level: Breakdown by workflow type (Extract, Transform, Load, Metadata)
• Stage Level: Individual stage analysis with optimization recommendations

Troubleshooting

Common Issues

1. Externally-managed-environment: Use virtual environment
2. Connection Errors: Verify Spark History Server URL and port
3. Authentication Issues: Ensure fresh cookies for browser mode
4. Permission Errors: Check file write permissions for local mode

Debug Mode

# Enable debug logging
spark-analyzer configure --debug
spark-analyzer analyze --debug

Support

For questions, issues, or technical support, contact Onehouse support at [e-mail].