A Canary Sizing Decision

Scenario

The problem

Every canary in Production Automation and the Capstone was exactly one device, chosen without ever explaining why one was the right number. For a 3-device fleet, one canary is a third of everything. For a 200-device fleet, one canary tells you almost nothing before the other 199 go out. This lab writes the sizing logic those modules skipped, then checks that evaluate_gate() -- unchanged -- still works correctly once the canary itself is more than a single device.

A decision exercise, not a new rollout pipeline

Nothing here connects to NetBox, Git, or an actual staged rollout against NAPALM. This lab is deliberately narrow: one sizing function, and one confirmation that the already-proven gate logic holds at a different canary size.

Task

Write recommend_canary_size(fleet_size, min_canary=1, min_percent=0.1, max_canary=5) that:

Computes min_percent of fleet_size, rounded up (a canary of "2.3 devices" isn't meaningful — round up to guarantee at least that much real coverage).
Returns at least min_canary, but never more than max_canary, even for a very large fleet.

Then, reusing evaluate_gate() exactly as Production Automation defined it (do not modify it), evaluate two 2-device canary scenarios:

One where both devices pass.
One where one device passes and one fails.

starter.py

import math

MIN_CANARY = 1
MIN_PERCENT = 0.1
MAX_CANARY = 5


def recommend_canary_size(fleet_size, min_canary=MIN_CANARY, min_percent=MIN_PERCENT, max_canary=MAX_CANARY):
  # Your code here.
  ...


# Reused verbatim from Production Automation -- do not modify this function.
def evaluate_gate(stage_result, threshold=1.0):
  total = len(stage_result)
  passed = sum(1 for r in stage_result.values() if r["applied"] and r["validated"])
  pass_rate = passed / total if total else 0.0
  return "PASS" if pass_rate >= threshold else "FAIL"

Starter code. evaluate_gate() is given to you unmodified -- your task is recommend_canary_size() plus applying evaluate_gate() to the two scenarios described above.

Hints

Expected output

expected output

fleet_size=3 -> recommended canary size: 1
fleet_size=20 -> recommended canary size: 2
fleet_size=100 -> recommended canary size: 5

2-device canary, both pass: PASS
2-device canary, one fails: FAIL

Real, verified output. A 3-device fleet gets the floor (1); a 20-device fleet gets a real 10% (2); a 100-device fleet hits the ceiling (5, not 10) -- and evaluate_gate() produces the correct decision either way, unmodified.

Validation checklist

recommend_canary_size(3) returns 1 (the floor, since 10% of 3 rounds up to 1 anyway)
recommend_canary_size(20) returns 2, not 1 or 3
recommend_canary_size(100) returns 5, not 10 -- the ceiling caps it
evaluate_gate() on a 2-device, both-pass canary returns 'PASS'
evaluate_gate() on a 2-device, one-fails canary returns 'FAIL'

Self-check only -- nothing here is submitted or graded. Compare against your own output manually.

Solution

Try the task yourself first -- the full solution and explanation stay hidden until you click above.

Key takeaways

A canary size of "one" was a choice Production Automation never explained -- this lab writes the reasoning that choice was standing in for.
Rounding up (math.ceil) and enforcing both a floor and a ceiling are what keep a percentage-based rule sensible at both very small and very large fleet sizes.
evaluate_gate() needed zero changes to work correctly against a multi-device canary -- its threshold logic was already general, just never exercised that way before.
No real fleet, device, or rollout was involved -- this is a sizing and decision exercise only.

A Canary Sizing Decision

Objective

Prerequisites

Scenario

Task

Expected output

Solution

Related lessons