Xray
1h 6h 24h 48h 3d 5d 7d 14d 30d ‹
Last 11719 seconds▾
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 19:03:31.
› →|
updated 1d 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.

At a glance 0

stored 32h ago all 0

Every service on master over the last 11719s, one small chart each, ordered by what moved most rather than by name.

How to read this page — small multiples

One card per service. The point of a grid like this is not to read any single chart closely — it is to spot the shape that is unlike the others, then click into it. Each card carries three numbers and a chart, and every one of them is about the same window as the range picker above.

the chart
One point per 5-minute bucket, left to right across the whole window. There is no y-axis on purpose: the height is scaled to this service's own range, so a tall spike means tall FOR THIS SERVICE, and two cards' heights are not comparable. Compare shapes, not heights.
bottom-right value
The median of the second half of the window — a representative recent value, not the last point and not the peak. Chosen because the last bucket is often partial and the peak is often one incident.
percentage, top-right
How the second half of the window compares with the first, using the MEDIAN of each half. +40% means the typical bucket in the recent half was 40% higher. Under ±10% it says `flat` rather than a number, because small movement is noise.
peak
Shown under the chart when the highest bucket is far above the median, so a single outlier spike is visible as a fact rather than distorting the summary.
none / flat card
Every service on the cluster is shown, including ones with nothing to plot for the selected measure — eight of them have Neo4j activity and no HTTP traffic at all, so `p95 latency` has genuinely nothing to draw. The card is flat and says which signal is absent, and its value is `—` rather than `0`: a service with no requests does not have a p95 of zero, it has no p95.
thin
Fewer than 15% of buckets had any traffic. The chart is still shown, the percentage is not — a trend from a handful of buckets is arithmetic, not a trend.

What this page will not tell you

  • Whether a value is GOOD. 250ms is fine for one service and terrible for another; this page has no objectives in it. Error budgets does.
  • Why something moved. A shape here is a prompt to open the service, not a cause — the change timeline is where deploys sit on the same axis.
  • Anything about a service with no traffic in the window: it is absent from the grid rather than shown as zero, because zero and no-data are different facts.

No service had traffic on master in this window.