Xray
1h 6h 24h 48h 3d 5d 7d 14d 30d ‹
Last 1 hour▾
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 09:33:01.
› →|
updated 22h 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.

Change timeline 0

pushed 23h ago 0 shown authoritative back 75d

Everything that changed on automation in the last 1h — what git deployed, what the cluster actually ran, and what someone changed by hand — with traffic, errors and latency for equal windows either side of each deploy. A deploy row is a commit in the config repo that ArgoCD applied; a ran row is the image a pod actually started, whose tag carries a commit from the service's own repo. Those are two different repositories, so each id says which it is and links to it. ?A deploy row compares the 60 minutes before the deploy with the 60 minutes after it, from the same rollups the charts read. Where the after-window is not complete yet the row says so rather than comparing unequal windows, which would manufacture a regression out of arithmetic.

How to read this page — change timeline

Three trails on one axis: what git deployed, what the cluster actually ran, and what someone changed by hand. They are kept apart rather than merged, because which kind a change was is usually the answer.

deploy
ArgoCD synced a commit for this app. This is what git ASKED for.
ran
A new pod template started — from the ReplicaSet trail, so it is what the cluster ACTUALLY ran. Usually right after a deploy; when it appears without one, something changed outside git.
outside git
A write straight to the cluster, from the audit log: a kubectl edit, a patched deployment, a deleted configmap. These are the ones worth reading first.
config <sha>
The commit in the config repo (ash-nazg) that ArgoCD applied. Links to that commit.
image / code
The container image that started, and the commit inside its tag — which is a commit in the SERVICE's own repo, a different repository from the config commit.
before → after
Traffic, p50, p95 and error rate for equal windows either side of the deploy, from the same rollups the charts read. Green is better, red worse, grey is inside ±5%.
too recent / too few samples
The comparison is refused rather than guessed: an after-window that is not complete yet, or fewer than 30 latency samples on one side.

What this page will not tell you

  • That a deploy CAUSED a change beside it. Both facts are true and adjacent; the causal claim is yours to make. Two deploys inside one window means two candidates.
  • Every change. ArgoCD keeps only its last 10 revisions per app, so anything older than this collector has been running is gone — the provenance chip says how far back the stored trail is authoritative.
  • Changes that produced no new pod template — a config value read at runtime, for instance — leave no `ran` row.

No recorded change on automation in this window.

That is a real answer, not a gap: the collector runs every 15 minutes and the trail is authoritative back 75d. Widen the range, or switch environment on the ArgoCD clusters page.