GitHub - dlt/faultline: Self-hosted embedded error tracking Rails engine. Track errors, get notified via Telegram/Slack/webhooks, and resolve issues with a clean dashboard—all without external services or SaaS fees.

A self-hosted error tracking engine for Rails 8+ applications. Track errors, get notified, and resolve issues—all without external services.

Features

• Automatic Error Capture - Rack middleware + Rails error reporting API
• Smart Grouping - Errors are grouped by fingerprint (class + message + location)
• Debugger Inspector - Side-by-side source code and local variables view, like a real debugger
• Local Variables Capture - Automatic capture of variable values at the raise point
• Full-Text Search - Search errors by exception class, message, or file path
• Status Management - Mark errors as resolved, unresolved, or ignored
• Auto-Reopen - Resolved errors automatically reopen when they recur
• GitHub Integration - Create GitHub issues directly from errors with full context
• Rate Limiting - Configurable cooldown prevents notification spam during error storms
• Pluggable Notifiers - Telegram, Slack, Email (ActionMailer or Resend), webhooks, or build your own
• Standalone Dashboard - Clean Tailwind UI with interactive charts and time-range zooming
• Configurable Authentication - Integrate with Devise, Warden, or custom auth
• Request Context - Capture URL, params, headers, user info, and custom data
• APM (Experimental) - Track response times, throughput, database query counts, span waterfall timelines, and CPU flame graphs per endpoint

Requirements

• Ruby >= 3.2
• Rails >= 8.0
• PostgreSQL, MySQL, or SQLite

Installation

Add to your Gemfile:

gem "faultline", git: "https://github.com/dlt/faultline.git"

# Or from RubyGems (when published)
# gem "faultline"

Run the installer:

bundle install
rails generate faultline:install
rails db:migrate

Configuration

Edit config/initializers/faultline.rb:

Authentication (Required for Production)

Faultline.configure do |config|
  # Devise with admin role
  config.authenticate_with = lambda { |request|
    user = request.env["warden"]&.user
    user&.admin?
  }

  # Or any authenticated user
  config.authenticate_with = lambda { |request|
    request.env["warden"]&.authenticated?
  }
end

GitHub Integration

Create GitHub issues directly from the error dashboard with full context including stack traces, local variables, and source code snippets.

# Store in credentials: rails credentials:edit
# faultline:
#   github:
#     token: "ghp_xxxxx"

config.github_repo = "your-org/your-repo"
config.github_token = Rails.application.credentials.dig(:faultline, :github, :token)
config.github_labels = ["bug", "faultline"]  # default labels for created issues

Notifications (Optional)

Notifications are optional. Faultline will track all errors and display them in the dashboard regardless of whether any notifiers are configured. Add notifiers only if you want to be alerted when errors occur.

Resend (Email)

# Store in credentials: rails credentials:edit
# faultline:
#   resend:
#     api_key: "re_xxxxx"

config.add_notifier(
  Faultline::Notifiers::Resend.new(
    api_key: Rails.application.credentials.dig(:faultline, :resend, :api_key),
    from: "errors@yourdomain.com",
    to: "team@example.com"  # or array: ["dev@example.com", "ops@example.com"]
  )
)

Email (ActionMailer)

Use your app's existing mail configuration (SMTP, SendGrid, Postmark, Mailgun, etc.) without any external API:

config.add_notifier(
  Faultline::Notifiers::Email.new(
    to: ["team@example.com", "oncall@example.com"],
    from: "errors@yourdomain.com"  # optional, defaults to ActionMailer default
  )
)

Emails are sent asynchronously via deliver_later, so Active Job must be configured. The email includes:

• Exception class and message
• Occurrence count and timestamps
• File location and stack trace
• User and request info (if available)

Telegram

# Store in credentials: rails credentials:edit
# faultline:
#   telegram:
#     bot_token: "your-bot-token"
#     chat_id: "your-chat-id"

config.add_notifier(
  Faultline::Notifiers::Telegram.new(
    bot_token: Rails.application.credentials.dig(:faultline, :telegram, :bot_token),
    chat_id: Rails.application.credentials.dig(:faultline, :telegram, :chat_id)
  )
)

Slack

config.add_notifier(
  Faultline::Notifiers::Slack.new(
    webhook_url: Rails.application.credentials.dig(:faultline, :slack, :webhook_url),
    channel: "#errors",
    username: "Faultline"
  )
)

Custom Webhook

config.add_notifier(
  Faultline::Notifiers::Webhook.new(
    url: "https://your-service.com/errors",
    method: :post,
    headers: { "Authorization" => "Bearer #{ENV['WEBHOOK_TOKEN']}" }
  )
)

Rate Limiting

Prevent notification spam during error storms:

config.notification_cooldown = 5.minutes  # default, nil to disable

Notification Rules

config.notification_rules = {
  on_first_occurrence: true,           # New error types
  on_reopen: true,                     # Resolved errors that recur
  on_threshold: [10, 50, 100, 500],    # At these occurrence counts
  critical_exceptions: [               # Always notify for these
    "Stripe::APIError",
    "ActiveRecord::StatementInvalid"
  ],
  notify_in_environments: ["production"]
}

Rails Error Reporting API

Optionally subscribe to Rails' error reporting API to capture errors from background jobs and explicit Rails.error calls:

config.register_error_subscriber = true

This captures errors reported via:

• Rails.error.handle { ... } - handled errors (swallowed)
• Rails.error.record { ... } - errors re-raised after reporting
• Rails.error.report(exception) - manual reporting

The middleware captures unhandled exceptions with local variable capture. The error subscriber captures everything else. Both can be enabled together.

Error Filtering

# Exceptions to ignore
config.ignored_exceptions = [
  "ActiveRecord::RecordNotFound",
  "ActionController::RoutingError"
]

# Bots/crawlers to ignore
config.ignored_user_agents = [/bot/i, /crawler/i, /Googlebot/i]

# Paths to ignore
config.middleware_ignore_paths = ["/assets", "/health"]

Custom Context

Add custom data to every error occurrence. The lambda receives the request and Rack env, and should return a hash:

config.custom_context = lambda { |request, env|
  controller = env["action_controller.instance"]
  {
    account_id: controller&.current_account&.id,
    tenant: request.subdomain,
    feature_flags: controller&.enabled_features
  }
}

This data is stored with each occurrence and displayed in the dashboard under "Custom Context".

Usage

Dashboard

Visit /faultline to access the error dashboard.

Manual Tracking

begin
  risky_operation
rescue => e
  Faultline.track(e, {
    request: request,
    user: current_user,
    custom_data: { order_id: @order.id }
  })
  raise
end

Local Variables Capture

Faultline automatically captures local variables at the point where exceptions are raised. This helps you debug errors without needing to reproduce them.

Variables are:

• Automatically captured via TracePoint when exceptions are raised
• Filtered for sensitive data (passwords, tokens, API keys, etc.)
• Serialized safely with depth limits and circular reference handling
• Displayed in the occurrence detail page

No configuration needed—it works out of the box.

Callbacks

# Skip certain errors
config.before_track = lambda { |exception, context|
  return false if exception.message.include?("Timeout")
  true
}

# Post-tracking integration
config.after_track = lambda { |error_group, occurrence|
  Analytics.track("error", { type: error_group.exception_class })
}

Custom Fingerprinting

# Group errors by custom criteria
config.custom_fingerprint = lambda { |exception, context|
  { extra_components: [context.dig(:custom_data, :tenant_id)] }
}

Building Custom Notifiers

class MyNotifier < Faultline::Notifiers::Base
  def initialize(api_key:)
    @api_key = api_key
  end

  def call(error_group, error_occurrence)
    data = format_message(error_group, error_occurrence)
    # Send to your service
    MyService.notify(data, api_key: @api_key)
  end

  # Optional: control when to notify
  def should_notify?(error_group, error_occurrence)
    error_group.occurrences_count < 100  # Stop after 100
  end
end

# Usage
config.add_notifier(MyNotifier.new(api_key: "..."))

Database Tables

The engine creates three tables for error tracking:

• faultline_error_groups - Grouped errors by fingerprint
• faultline_error_occurrences - Individual error instances
• faultline_error_contexts - Custom key-value context data

And two tables for APM (if enabled):

• faultline_request_traces - Individual request performance data, including span JSON
• faultline_request_profiles - CPU profile data for flame graph visualization (when profiling is enabled)

Data Retention

Configure automatic cleanup:

config.retention_days = 90  # nil = keep forever

Set up a scheduled job to clean old data:

# In a cron job or Sidekiq scheduler
Faultline::ErrorOccurrence
  .where("created_at < ?", 90.days.ago)
  .in_batches
  .delete_all

APM (Application Performance Monitoring)

Experimental: APM support is new and may have rough edges. The schema, configuration options, and dashboard are subject to change in future releases. Feedback welcome.

Faultline includes basic APM to track request performance alongside your errors. It's lightweight, automatic, and requires no code changes.

What It Tracks

How It Works

APM hooks into Rails' built-in ActiveSupport::Notifications:

Your Rails App (no changes needed)
         │
         ▼
Rails fires "process_action.action_controller" after each request
Rails fires "sql.active_record" for each SQL query
         │
         ▼
Faultline::Apm::Collector listens and stores metrics
         │
         ▼
Dashboard at /faultline/performance

No manual instrumentation required—Rails already emits these events for every request.

Enabling APM

APM is opt-in and disabled by default:

# config/initializers/faultline.rb
Faultline.configure do |config|
  config.enable_apm = true
end

Then run migrations (if you haven't already):

Visit /faultline/performance to see the dashboard.

Configuration Options

Faultline.configure do |config|
  # Enable APM (default: false)
  config.enable_apm = true

  # Sample rate: 1.0 = 100%, 0.1 = 10% (default: 1.0)
  # Use sampling for high-traffic apps to reduce storage/overhead
  config.apm_sample_rate = 1.0

  # Paths to ignore (default: falls back to middleware_ignore_paths)
  # Faultline's own routes are always ignored
  config.apm_ignore_paths = ["/health", "/assets", "/up"]

  # Data retention in days (default: 30)
  config.apm_retention_days = 30

  # Capture individual operation spans (SQL, HTTP, Redis, views) — default: true
  # Adds a waterfall timeline to each request trace detail page
  # Capped at 500 spans per request to limit memory use
  config.apm_capture_spans = true

  # CPU profiling via stackprof (default: false)
  # Requires the stackprof gem: gem "stackprof"
  config.apm_enable_profiling = false
  config.apm_profile_sample_rate = 0.1   # profile 10% of requests
  config.apm_profile_interval = 1000     # microseconds between samples
  config.apm_profile_mode = :cpu         # :cpu, :wall, or :object
end

Dashboard Features

Index page (/faultline/performance):

• Summary stats: total requests, avg response time, p95, error rate
• Response time chart (avg/min/max over time)
• Throughput chart (requests per time bucket)
• Slowest endpoints table with avg, p95, request count, error rate
• Time period selector: 1h, 6h, 24h, 7d, 30d

Endpoint detail page (/faultline/performance/UsersController%23show):

• Stats for a single endpoint
• Response time chart for that endpoint
• Individual request traces (sortable, searchable, paginated)

Trace detail page (/faultline/traces/:id):

• Full timing breakdown for a single request
• Waterfall timeline of SQL, HTTP, Redis, and view spans (when apm_capture_spans is enabled)
• Embedded flame graph viewer powered by speedscope (when profiling is enabled and the request was sampled)

Data Cleanup

APM data is cleaned up separately from error data:

# Run manually
rails faultline:apm:cleanup

# Or schedule it (cron, Sidekiq, etc.)
Faultline::RequestTrace.cleanup!  # Uses apm_retention_days
Faultline::RequestTrace.cleanup!(before: 7.days.ago)  # Custom

Performance Overhead

The overhead is:

• ~0.01ms per SQL query for counting
• ~1-5ms per request for the database INSERT (after response is sent)

The INSERT happens after Rails sends the response, so users don't wait for it.

Caveats & Limitations

1. Not a full APM replacement — This is basic request-level monitoring. It doesn't provide:

   • Distributed tracing across services
   • Memory profiling or GC introspection
   • N+1 query detection
   • Advanced alerting on performance metrics

2. Dashboard queries can be slow at scale — With millions of traces, aggregate queries take longer. Mitigations:

   • Use shorter time periods (1h, 6h instead of 30d)
   • Use sampling to reduce data volume
   • Dashboard results are cached for 1 minute

3. P95 calculation — PostgreSQL uses native PERCENTILE_CONT (fast). SQLite/MySQL fall back to ORDER BY + OFFSET (slower on large datasets).

4. Storage grows with traffic — At 100 req/sec with 100% sampling:

   • ~8.6M rows/day
   • ~260M rows/month
   • Use apm_retention_days and sampling to manage this

5. No alerting — APM doesn't have its own alerting. Use error tracking notifications for 5xx responses, or build custom alerts from the data.

When to Use Faultline APM

Good fit:

• Rails monoliths wanting basic performance visibility
• Teams already using Faultline for errors
• Projects avoiding external APM services (cost, privacy)
• Development/staging environments

Consider dedicated APM if you need:

• Distributed tracing (microservices)
• Detailed profiling (memory, CPU, GC)
• External service monitoring (Redis, Elasticsearch, HTTP)
• Advanced alerting on performance metrics
• Historical data over months/years

Comparison with Alternatives

Faultline is ideal for:

• Teams wanting full control over error data
• Projects with privacy/compliance requirements (GDPR, HIPAA)
• Rails-only applications without complex frontend needs
• Developers who prefer simplicity over feature bloat
• Organizations avoiding vendor lock-in and recurring SaaS costs

Development

cd engines/faultline
bundle install
bundle exec rspec

License

MIT License. See LICENSE for details.