Watchdog Process Manager

• Long-running watcher processes need supervision
• Automatic restart of failed orchestrator processes is required
• Service availability and uptime must be maintained
• Process health and performance need monitoring
• Alerting is needed for process failures or anomalies
• Resource usage needs to be tracked and controlled

Process Supervision

Restart Policies

The watchdog supports various restart policies:

• Always: Always restart failed processes (default)
• On-Failure: Only restart if process exits with non-zero code
• Never: Don't restart, just log the failure
• Exponential Backoff: Increase delay between restart attempts

Health Checks

Processes can be monitored using:

• Heartbeat: Regular signals from the process
• Resource Usage: CPU/memory thresholds
• External Ping: Network-based health checks
• File-based: Checking for specific file states

Configuration

Process Definitions

Processes are defined with configuration specifying:

• Command to execute
• Restart policy
• Environment variables
• Working directory
• Resource limits
• Health check parameters

Example Configuration