zeekay
1b367e2b62
The tree was a debranded snapshot of upstream taken around 2026-02-20, and upstream has moved about five months since. This replaces it with upstream master, mechanically re-derived: bin/debrand does the rename, and the pipeline and squashed migrations are restored from the shipped tree. Not a merge. The previous import replaced the tree wholesale in a single commit, so there is no shared content history to merge against. Reproducing the rename on the new tree is what actually converges. What survives: .hanzo/workflows, the Dockerfiles and bin/docker* entrypoints, compose files, LLM.md, and insights/migrations. The squashed initial migration is load-bearing — the live database records it by name, so upstream's ~918 individual migrations cannot come back. Nine products are dropped as unreferenced. Twenty-one more that we do not run are kept anyway because the core app still imports them; unpicking those is a refactor and does not belong in a rebase. This base does not build yet, and the reasons are not in the rename: upstream now wants Python 3.13.13 against our 3.12.12 and Django 5.2 against our 4.2, and the dependency manifests still name upstream packages that we republish under our own names. Those are decisions, not renames.
1.8 KiB
Characterize how a metric behaves in a suspicious window compared to a healthy baseline. This is the FIRST call to make when investigating "metric X is rising/dropping — why?": one call answers how big the change is, exactly when it started, and which label values moved.
Required: metricName (exact — discover via metric-names-list first) and anomalyFrom (when things started looking wrong; the alert fire time works). anomalyTo defaults to now. The baseline defaults to the equal-length window immediately before anomalyFrom; pass baselineFrom/baselineTo to compare against a known-good period instead (e.g. same time yesterday).
The aggregation is auto-picked from the metric's OTel type (counter -> rate, gauge -> avg, histogram -> histogram_quantile 0.95); override with aggregation if you need a different lens. Use filters to scope to a service or region. candidateKeys controls which label keys are drilled into for the movers analysis — omit it to auto-discover.
Read the report in this order:
direction+change_ratio+anomaly_peak: how bad is it? (flatmeans the windows don't meaningfully differ — widen the anomaly window or check you have the right metric.)onset_time: when it started. Use this exact timestamp as the pivot for cross-signal correlation.top_movers: which label values changed. One label value moving alone (e.g. a single pod or shard) points at a localized culprit; everything moving together points at a shared cause upstream.series: the full baseline+anomaly time series if you need to eyeball the shape.
Then correlate: query logs (query-logs with the same service and a window around onset_time, severity error first) and traces (APM span tools, same service/window) to find what broke and its blast radius. All parameters are nested inside a query object.