From 601c9ece3f0ce24792d1dc1f9dce862a6826e876 Mon Sep 17 00:00:00 2001 From: ZverGuy Date: Thu, 5 Nov 2026 00:18:47 +0700 Subject: [PATCH] docs(dashboard): note why the metrics panels refresh on an interval The console rule against refetchInterval assumes a watch is available, which holds for Kubernetes lists but not for the proxy's Prometheus query API. Record the exception so the metrics panels' refresh does not read as a rule violation. Assisted-by: LLM Signed-off-by: ZverGuy --- packages/system/dashboard/images/console/CLAUDE.md | 1 + 1 file changed, 1 insertion(+) diff --git a/packages/system/dashboard/images/console/CLAUDE.md b/packages/system/dashboard/images/console/CLAUDE.md index 96419d118..036e527d6 100644 --- a/packages/system/dashboard/images/console/CLAUDE.md +++ b/packages/system/dashboard/images/console/CLAUDE.md @@ -81,6 +81,7 @@ the apiserver checks that user's RBAC. to a chunked-encoding WATCH against the same `resourceVersion`. Don't poll. Don't add `refetchInterval`. If you need a one-shot, pass `{ watch: false }`. + The one exception is the metrics panels in `components/observability`: the observability proxy serves the Prometheus query API, which has no watch, so they refresh on a fixed interval instead. Log lines stream through the tail endpoint and are not polled. 4. **Tenant scoping.** Most resources live in `tenant-` namespaces. Pull the active tenant from `useTenantContext()` — never read the namespace from a URL param or guess it.