SLOs and Apdex
Say how much of a service's traffic must succeed or be fast enough, watch the error budget and its burn rate, and set Apdex targets.
An SLO (service level objective) says how much of a service's or an endpoint's requests must be good over 7 or 30 days. WizeLens shows how much error budget is left and how fast it burns, and an alert can watch the burn rate. SLOs run on the local hub and on WizeLens Cloud. Open SLOs in the menu and choose New SLO.
An SLO
| Setting | What it does |
|---|---|
| Name (optional) | What the list and alerts call it |
| Service | The service it covers |
| Endpoint (optional) | One endpoint, such as GET /api/orders/{id}, or all of them |
| Good means | Requests succeed, or Requests are fast enough |
| Within (ms) | For fast enough: a good request finishes in this time |
| Target (%) | The share that must be good, such as 99.9 |
| Window | Over 30 days or Over 7 days, ending now |
The goal reads as a sentence: "99.9% of checkout requests succeed, over 30 days".
Error budget and burn rate
The error budget is the share of requests that may be bad. At 99.9% over 30 days, 0.1% of that window's requests may fail. Budget Left is how much of it is not spent yet: 100% when nothing went wrong, 0% once it is spent.
The burn rate is how fast the budget goes: the bad share now, over the bad share allowed. At 1 the budget lasts exactly the window. At 14.4 for an hour, 2% of a 30-day budget is gone: page someone. At 6 for six hours, open a ticket.
The SLOs page
Objectives lists each SLO:
| Column | What it shows |
|---|---|
| SLO | Its name |
| Goal | The sentence above |
| Now | The good share over the window, red when under the target |
| Budget Left | The error budget not spent yet |
| Burn (1 h / 6 h) | The burn rate over the last hour and the last 6 hours |
| Requests | Requests in the window |
Each SLO has its own page: its goal, the good share now, its requests, the burn now, and a link to the service or endpoint it covers. Two charts show Error Budget Left (0% once spent) and Good Requests against the target over the window. Edit and Delete change it.
Alert on burn rate
Alert on burn rate on an SLO's page opens a new alert rule that fires when the burn rate goes above 14.4 over 1 hour. Change the rate and the window to fit. The notice reads:
SLO Checkout availability burns its budget 16.2× as fast as allowed (above 14.4×)Apdex
Apdex scores how fast requests felt, from 0 to 1, against a target time:
- Satisfied: at or under the target.
- Tolerating: under four times the target.
- Frustrated: slower than that, or failed.
The score is satisfied plus half of tolerating, over all requests.
| Score | Rating |
|---|---|
| 0.94 and up | Excellent |
| 0.85 to 0.94 | Good |
| 0.70 to 0.85 | Fair |
| 0.50 to 0.70 | Poor |
| Under 0.50 | Unacceptable |
The Apdex card on the Overview and on each service's page shows the score, its rating, and the target. On the hub and the cloud, set each service's target under Apdex Targets on the SLOs page, then Save targets. An empty target uses 500 ms, and your app's own dashboard uses 500 ms for every service. With several targets, the Overview scores each service against its own.
Teams and projects
- In the cloud, members, admins, and owners make and change SLOs and targets; viewers see them.
- In the cloud, under All projects, the list shows each SLO's project, a new SLO asks which project it belongs to, and Apdex targets are set per project: pick one in the menu.
- Ctrl K finds SLOs by name.