Xray
hourly rollups 1h 6h 24h 48h 3d 5d 7d 14d 30d ‹
Last 24 hours▾
Ends now. Anything up to this cluster's 30-day retention.
Start and end times
Cancel
Interpreted in UTC (change it at the right of this bar). Earliest available: 2026-10-19 08:00:53.
› →|
updated 21h ago
Sample data. This database was generated by scripts/seed_dev_db.py — every service, deploy, person and figure on this page is synthetic. Names are sample.*@seed.invalid and image tags are prefixed SEED-. Nothing here happened.

Error budgets 17

stored 21h ago all 17objectives derived from history

Every service on master over the last 24h, against an objective — with how much of the budget is gone and how fast it is going. ?A budget is the failures an objective permits. Burn rate is how fast you are spending it: 1 spends exactly the window's budget over the window, 14.4 spends 2% of a 30-day budget in an hour. Two windows are shown because they answer different questions — the 1h catches something breaking now, the 6h catches a slow leak the 1h would dismiss as noise.

How to read this page — error budgets

An objective turns a latency or error number into a verdict, and a burn rate turns the verdict into a schedule: not "we are at 99.3%" but "at this rate the budget is gone by Thursday".

objective
The share of requests allowed to fail. `der` means DERIVED from this service's own history — see the banner — and `dec` would mean a declared target, which nothing in the fleet has yet.
achieved
The share that actually succeeded in this window. Red when it is under the objective.
budget left
How much of the permitted failure is unspent. 100% means nothing has failed; 0% means the objective is already missed for this window.
burn 1h / 6h
How fast the budget is being spent, relative to spending it exactly over the window. 1x is on-pace; 14.4x spends 2% of a 30-day budget in an hour. Two windows, because the 1h catches something breaking now and the 6h catches a slow leak the 1h dismisses as noise.
n/a
Fewer than 20 requests in that window. A burn rate from ten calls is noise, so none is shown.
bad
5xx responses, plus — in `availability + slow` mode — requests slower than 1.5x this service's own baseline p95. 4xx is deliberately excluded: a client asking for something that does not exist is not the service failing.

What this page will not tell you

  • Whether users are happy. A derived objective measures a service against its own past, not against what its callers need. Only a declared target can do that.
  • Anything about a service with no traffic: it is absent rather than shown as perfect.
  • A comparison between services. Two services with different derived objectives are being held to different bars, on purpose.
These objectives are derived, not agreed. Nothing in the fleet declares a target, so each service is measured against twice its own normal failure rate over the last 7 days. That answers “is this worse than this service usually is”, which is usually what people mean — but it is not a promise to anyone, and a service that has always been unreliable gets an unambitious objective. Declaring real targets is an upstream change; when they exist they replace this and the column below will say declared.
Columns
Views
Export CSV
Error budget, burn rate and verdict per service on master
Service Verdict Objective Achieved Budget left Burn 1h Burn 6h Calls Bad p95 p99
docsvc below objective 99.5%der 99.42% 0% n/a n/a 1k 6 1.56 s 4.47 s
chat-agent healthy 99.9%der 100.00% 100% n/a n/a 30 0 57.00 s 1.8 min
dashql healthy 99.5%der 100.00% 100% n/a n/a 103 0 4.52 s 4.90 s
docproc healthy 99.0%der 100.00% 100% n/a n/a 222 0 420 ms 963 ms
enrollsvc healthy 99.9%der 100.00% 100% n/a n/a 301 0 389 ms 484 ms
execsvc healthy 95.0%der 100.00% 100% n/a n/a 825 0 861 ms 1.89 s
feedsvc healthy 99.9%der 100.00% 100% n/a n/a 1k 0 512 ms 1.21 s
notifsvc healthy 99.9%der 100.00% 100% n/a n/a 203 0 — —
onboardingsvc healthy 99.9%der 100.00% 100% n/a n/a 3k 0 126 ms 434 ms
ontologysvc healthy 99.9%der 100.00% 100% n/a n/a 22 0 417 ms 484 ms
patientsvc healthy 99.9%der 100.00% 100% n/a n/a 90 0 1.94 s 4.32 s
regsvc healthy 99.9%der 100.00% 100% n/a n/a 6 0 190 ms 198 ms
schedsvc healthy 99.9%der 100.00% 100% n/a n/a 23 0 1.92 s 4.31 s
studiosvc healthy 99.0%der 100.00% 100% n/a n/a 2k 0 26.0 min 90.1 min
tusker healthy 99.9%der 100.00% 100% n/a n/a 61 0 886 ms 1.39 s
virtual-assistant healthy 99.9%der 100.00% 100% n/a n/a 531 0 1.58 s 15.20 s
workplansvc healthy 99.9%der 100.00% 100% n/a n/a 44 0 1.73 s 1.95 s