wu-sheng opened a new pull request, #65: URL: https://github.com/apache/skywalking-horizon-ui/pull/65
## Why Three independent scale / resilience gaps on the topology maps and the 3D Infra Map, surfaced during RC0 review: - A very large layer graph would try to lay out thousands of nodes — an unreadable map (and a slow paint) instead of a clear "narrow your scope" message. - The 3D Infra Map fetched per-node metric values one chunk at a time, so load rings / traffic values filled in slowly on large layers. - When a metric batch failed (a transient OAP error), blank cells were indistinguishable from a real zero — and on the endpoint-dependency map nodes could silently vanish — with no signal to the operator. ## What - **3D Infra Map — parallel metric load.** Per-node metric values now fetch in bounded-concurrency batches (new `metricConcurrency` config, default 4, max 8) instead of serially. Output-identical; each request still carries ≤12 services. - **Topology — reject oversized graphs with guidance.** When a layer graph exceeds 5000 services or 15000 calls, the service-map route returns a `tooLarge` envelope (with live counts) and the UI shows a "too large to render" banner pointing at the focus picker / depth control (or the Topology tab in embedded mode). No truncation — the operator narrows scope and sees a complete map. - **All four maps — partial-metric advisory.** Service map, instance topology, deployment, and endpoint-dependency thread per-chunk soft-fail counts and emit `metricsPartial`; the UI shows a banner clarifying blank values may be unavailable, not zero. The endpoint-dependency copy instead warns some endpoints or links may be missing, since that route prunes valueless nodes. New strings translated across all 8 locales. ## Validation - Full preflight gate green: `pnpm -r type-check`, UI + BFF build, `pnpm -r lint`, **122 BFF + 113 UI unit tests**, `license:check` (0 invalid), BFF template `i18n:validate` (no findings). - i18n: 2 new keys registered in `en.json` and translated in all 7 other locales (drift check reports all catalogs complete). - Live against the demo OAP: `GENERAL` topology renders 6 nodes / 5 calls with no false `tooLarge` / `metricsPartial` banners on a healthy graph. - Reviewed by an adversarial multi-agent pass: the parallel 3D load and the four-route soft-fail are output-identical on success and strictly better on partial failure (per-chunk soft-fail vs. the prior break-on-first-error). -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
