Max Restarts

Prevent infinite restart loops by limiting how many times a supervisor will restart a failing agent.

Configuration

naos supervisor create main --strategy one-for-one --max-restarts 3

supervisors:
  main:
    strategy: one-for-one
    maxRestarts: 3
    restartWindow: 60    # seconds

How It Works

The supervisor counts restarts within the restart window:

Time 0s:  Agent fails → Restart 1/3
Time 10s: Agent fails → Restart 2/3
Time 20s: Agent fails → Restart 3/3 (max reached)
Time 25s: Agent fails → NOT restarted, marked as failed

Restart Window

The restart window determines the time period for counting restarts. After the window expires, the count resets.

Window: 60 seconds, Max: 3

Time 0s:  Agent fails → Restart 1/3
Time 10s: Agent fails → Restart 2/3
Time 70s: Agent fails → Restart 1/3 (window reset!)

Defaults

Setting Default
maxRestarts 3
restartWindow 60 seconds

What Happens at Max

When max restarts is reached:

  1. The agent is marked as failed
  2. The supervisor logs the event
  3. The agent is NOT restarted again
  4. An audit entry is created
naos supervisor status main

# Children:
#   worker-1   ● running    0 restarts
#   worker-2   ✗ failed     3 restarts (max reached)