Day 44 results: the error budget, in minutes and requests

Measured on: (machine, OS, Python version)

Output

Paste part 1, part 2, part 3 and the scoreboard here.

One sentence per prediction

P1 (budget in minutes for 99.9%):
P2 (budget in requests for the month):
P3 (incident-day burn rate):
P4 (combined availability of 5 dependencies):

The SLI vs the SLO

What availability did you measure, and were you meeting the 99.9% SLO? How much room was there between the SLI and the SLO:

The budget, and the burn

The 99.9% budget was __ failures / __ minutes a month. The incident day burned __x the sustainable rate, which would empty the month’s budget in __ days. What that tells you that a raw availability percentage does not:

Why not 100%

In one or two lines: what did the “each nine costs 10x” table show you, and why does a 100% target leave no room to ship:

Dependencies multiply

Five dependencies at 99.9% put your ceiling at __%, about __ minutes a month, before your own bugs. Which lever would you pull first to raise it:

The number that matters

A 99.9% SLO allows ___ minutes of downtime a month. Write the one sentence you would say in a design review to argue against “we need 100% uptime”: