Error budget

Work out the numbers for the SLO in "$ARGUMENTS", as references/slos-and-alerting.md describes.

Compute with code. Perform every calculation (totals, balances, ratios, variances, projections) with a script — Bash, Python or a spreadsheet formula written to a file — and include the computation alongside the result. Never do arithmetic mentally.

Prerequisites

The SLO target (for example 99.9%) and window (for example 30 days); optionally the request volume, a current burn rate, and how much of the budget is already consumed in this window (default: none, a full budget). Ask once for any that are missing.

Process

1. Compute with a script, shown in full: the budget as 1 − target; the budget in minutes over the window; the budget in failed requests when a volume is given.

2. The alert thresholds: for each standard window pair (14.4× over 1 hour and 5 minutes, page; 6× over 6 hours and 30 minutes, page; 1× over 3 days and 6 hours, ticket), the error-rate threshold (burn rate × (1 − target)) and the budget consumed when it fires (burn rate × long window ÷ window).

3. Time to exhaustion at the current burn rate when given: remaining budget × window ÷ burn rate, where the remaining budget is 1 minus the share already consumed; when no consumption was given, say the figure is "from a full budget".

4. Thresholds that can never fire: if burn rate × (1 − target) is 100% or more, that alert can never fire (the error rate cannot exceed 100%); say so, and use a lower burn rate or a longer window for it.

5. Without a way to run code, give the formulas with the user's values substituted, and label the results not computed.

Output

In chat: the budget, a table of the alert thresholds and budget consumed, the time to exhaustion, and the script. This skill writes no file.