Skip to main content

SLO and Alert Lab

Active use only. At the end of this lab you should have one SLO document and at least one working burn-rate alert in your capstone, not a theoretical discussion of either.

Retrieval Prompts​

  1. State the difference between an SLI, an SLO, and an SLA in your own words.
  2. Write the formula for an availability SLI as a ratio of events.
  3. For a 99.5% SLO over 30 days, how much is the error budget as a percentage of total events?
  4. State the two windows typically combined in a multi-window burn-rate fast-burn alert.
  5. Why is "CPU > 80%" usually a bad primary alert?

Compare and Distinguish​

Separate these pairs clearly:

  • SLI vs a "system metric" like CPU utilization
  • percent-based SLO vs event-based error budget
  • single-window alert vs multi-window burn-rate alert
  • page-worthy symptom vs ticket-worthy symptom
  • aspirational SLO vs defensible SLO for current architecture

Common Mistake Check​

For each statement, identify the error:

  1. "Our SLO is 100% -- anything less and users complain."
  2. "We alert if error rate exceeds 1%; that's our SLO alert."
  3. "Budget is fine; we're at 78% consumed with 3 days left in the window."
  4. "We used the AmazonFreeTierMetrics default for our SLO target."
  5. "The alert fires on CPU > 90% because that's when things get slow."

Mini Application​

For your capstone:

  1. SLI formula (write it as a ratio):

    SLI = <good events expression> / <total events expression>

  2. SLO + window + error budget (fill in concrete numbers):

    • target: ____ %
    • window: rolling ____ days
    • error budget (% of events): ____
    • error budget (absolute events, at current traffic): ____

  3. Consequence for missing the SLO (one sentence):

  4. Fast-burn alert (pseudocode):

    condition:  error_ratio(last ___) > <multiplier> * <budget fraction>
    AND         error_ratio(last ___) > <multiplier> * <budget fraction>
    action:     PAGE

  5. Slow-burn alert (pseudocode): same structure, looser thresholds, ticket not page.

Evidence Check​

This lab is complete only if:

  • library/raw/slo.md exists and contains the SLI, SLO, window, budget, and consequence
  • library/raw/error-budget-policy.md contains the 5-tier ladder
  • at least one burn-rate alert is configured in your monitoring tool
  • you have deleted or demoted to ticket status at least one non-SLO page-level alert
  • you can explain every threshold number in both alerts without checking a book