Resources / SLO
SLO
A service level objective: how often a product must be good, measured over a window, with the error budget and burn rates computed from samples (spec 0047 section 6).
Overview
| API | platform.etorm.com/v1alpha1 |
| Resource | slos (cluster-scoped inside its cluster) |
| Short names | slo |
| Available in | the platform and project clusters |
| Status | written by a controller through /status |
Minimal example
apiVersion: platform.etorm.com/v1alpha1
kind: SLO
metadata:
name: example
spec:
targetPercent: 50
indicator: {}
Spec
| Field | Type | Required | Default | Allowed | Description |
|---|---|---|---|---|---|
targetPercent | number | yes | 50 to 99.999 | The share of samples that must be good over the window, in percent (for example 99.9). | |
windowSeconds | integer | 2592000 | 3600 to 7776000 | The window the objective is measured over (default 30 days). | |
sampleSeconds | integer | 300 | 60 to 3600 | How often the indicator is sampled. Each sample is one write, so the default is five minutes. | |
fastBurn | number | 14.4 | Burn rate over the last hour at which the objective counts as burning. | ||
slowBurn | number | 6 | Burn rate over the last six hours at which the objective counts as burning. | ||
indicator | object | yes | What is sampled: exactly one of an Intent's Ready condition or a public probe. | ||
indicator.intentRef | object | An Intent in the same cluster: a sample is good when it is Ready. | |||
indicator.intentRef.name | string | yes | Name of the Intent. | ||
indicator.probe | object | A public URL: a sample is good when it answers as expected. | |||
indicator.probe.url | string | yes | up to 512 characters | The address to request with GET. | |
indicator.probe.expectCode | integer | 200 | The HTTP status of a good answer. | ||
indicator.probe.expectText | string | up to 256 characters | Text a good answer contains. |
Status
| Field | Type | Required | Default | Allowed | Description |
|---|---|---|---|---|---|
phase | string (one of) | Pending, Ready, Burning, Exhausted | Pending before the first sample, Ready while the budget lasts and no burn alert fires, Burning while one fires, Exhausted when the budget is spent. | ||
state | string (one of) | Healthy, Burning, Exhausted, Unknown | Healthy while the budget lasts and no burn alert fires; Burning while one fires; Exhausted when the budget is spent. | ||
good | integer | Good samples in the window. | |||
total | integer | Samples in the window. | |||
budgetRemaining | number | Fraction of the error budget left, as a share of the whole window (1 is all of it, 0 none, negative overspent). | |||
burnRate1h | number | How many times faster than allowed the budget burned over the last hour. | |||
burnRate6h | number | The same over the last six hours. | |||
lastSampleAt | integer | Unix time of the newest sample. | |||
recent | string | The last 72 samples, oldest first, as a string of 1 (good) and 0 (bad). | |||
hourly | list of object | Samples per hour over the window, oldest first (bounded by the window). | |||
hourly[].hour | integer | Unix time of the start of the hour. | |||
hourly[].good | integer | Good samples in the hour. | |||
hourly[].total | integer | Samples in the hour. | |||
conditions | list of object | Ready (budget left, not burning) and Degraded (burning), spec 0047 section 2. | |||
conditions[].type | string | Ready or Degraded. | |||
conditions[].status | string | True, False or Unknown. | |||
conditions[].reason | string | A word from the reason registry. | |||
conditions[].message | string | A sentence. |
Columns in listings
| Column | From |
|---|---|
| Target | .spec.targetPercent |
| Budget-Left | .status.budgetRemaining |
| Burn-1h | .status.burnRate1h |
| State | .status.state |
| Phase | .status.phase |