distributed systems / LEVEL 4

The release that spent a month in ten minutes

EST.40 MIN

THEORY / RETRIEVAL

What to restore

Calculate an integer error budget from request volume
Express observed consumption as milli-burn rate
Use multi-window burn evidence to pause progressive delivery

An SLO creates a finite budget for unreliability

At 99.9% availability, roughly 0.1% of valid events may be bad over the objective window. Burn rate compares observed consumption pace with that budget. A rate far above one can spend the whole window quickly even when a short dashboard percentage still looks close to target.

Alerts should predict budget exhaustion

Multi-window burn alerts combine a fast signal for severe incidents with a slower confirmation that reduces noise. Progressive delivery can pause or roll back on these signals, but the SLI denominator, excluded traffic and measurement gaps must be explicitly defined.

CHECKPOINT

What does a sustained burn rate of 10 mean for an error budget?

Confidence: 3/5

02 / IMPLEMENTATION

Implement the contract

Assess one request-count window. availability_target_ppm is in 1..999999, bad_requests cannot exceed total_requests, and allowed_bad is floor(total × (1_000_000-target)/1_000_000). Return remaining bad events and burn_milli; when allowed_bad is zero, any bad event saturates burn_milli.

Initializing editor…

CLOUD SANDBOXnetwork off · 256 MB · 12 s

1 / 64 KB

OUTPUT

Runner is waiting for a submission.