DrP: Meta’s Root Cause Analysis Platform at Scale

What It Is

Meta’s DrP is a root cause analysis (RCA) platform designed to automate incident investigation for large-scale systems, currently running 50,000 analyses daily. It moves beyond manual, error-prone investigations reliant on playbooks and scripts towards a programmatic, scalable solution for faster incident resolution.

DrP provides an expressive SDK for creating investigation workflows (“analyzers”), a scalable backend for execution, and integration with alerting and incident management tools, ultimately reducing on-call toil.

Why This Matters

Traditional incident investigation often struggles with the complexity of modern systems, leading to prolonged outages and significant engineering costs. Manual analysis is slow, inconsistent, and doesn’t scale, while automated systems often lack the flexibility to handle novel issues. DrP addresses this by providing a balance between automation and engineer control, resulting in a 20-80% reduction in MTTR.

Key Insights

• 50,000 analyses daily: DrP executes this volume across Meta’s infrastructure, 2025.
• SDK-driven automation: Analyzers, built with the DrP SDK, codify investigation workflows for consistency and repeatability.
• AI4Ops vision: DrP is evolving into an AI-native platform to enhance investigation accuracy and automation.

Working Example

# Example Analyzer (Conceptual - based on description)
def analyze_incident(alert_data):
    """
    Analyzes an incident based on provided alert data.
    """
    # Access data using DrP SDK libraries (e.g., time series correlation)
    correlated_events = correlate_events(alert_data.timestamp)

    # Perform anomaly detection
    anomalous_metrics = detect_anomalies(alert_data.metrics)

    # Isolate potential root cause
    root_cause = isolate_cause(correlated_events, anomalous_metrics)

    return root_cause

# Placeholder functions - replace with actual DrP SDK calls
def correlate_events(timestamp):
    # Returns a list of correlated events around the given timestamp
    pass

def detect_anomalies(metrics):
    # Returns a list of anomalous metrics
    pass

def isolate_cause(events, metrics):
    # Returns the likely root cause based on events and metrics
    pass

Practical Applications

• Meta’s Infrastructure: DrP is used across hundreds of teams to automate the investigation of incidents, improving system reliability.
• Pitfall: Relying solely on pre-defined playbooks without the flexibility to adapt to new incident types can lead to incomplete or inaccurate root cause analysis.

References:

• https://engineering.fb.com/2025/12/19/data-infrastructure/drp-metas-root-cause-analysis-platform-at-scale/