0 Fork 0

Files

T

matsandClaude Sonnet 5.5 1a536e7d0e Split the console from the resolver: fancy-dns-console writes the cluster

fancy-dns-console is a second binary holding the web UI, authentication and
every edit. It writes ClientGroup, BlockList and DNSZone objects through the new
kubeedit package and forwards all other /api routes to a running resolver
with its own key, dropping the browser's credentials, so it can run in the
cluster or on a workstation with a kubeconfig and nothing about the resolver's
placement matters. The resolver never imports kubeedit and the console is not on
the DNS path.

kubeedit validates every spec with the runtime's own converters
(kubesource.CheckGroup/CheckList/CheckZone) before writing, so the console
cannot store an object the resolver would reject. Zone edits keep the
file-based writer's behaviour: duplicate and below-a-delegation records are
refused, a delegation in a new zone also gives the zone its servers, and a
zone with no apex NS is never stored. Each object records the editing user in
an annotation; updates retry on conflict.

The resolver reports groups from ClientGroup objects as source "kubernetes" and
counter-only rows (the "(default)" bucket, or a group removed since) as
"history", and lists DNSZone zones as the editable ones. The console uses that
to leave the YAML's groups alone and to drop history rows, which the UI would
otherwise send back on save. Its GET /api/clients overlays the cluster's
current objects on the resolver's list, because the resolver only sees an edit
after its watch fires and a save made just after another would delete the
first.

An end-to-end test runs both halves against one fake apiserver: a group saved
through the console refuses queries, a zone built record by record answers
authoritatively, and a request with the resolver's own key is rejected by the
console.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

2026-10-16 08:43:41 +02:00

.codesearch

initial commit

2026-09-27 11:20:34 +00:00

public

initial commit

2026-09-27 11:20:34 +00:00

src

Split the console from the resolver: fancy-dns-console writes the cluster

2026-10-16 08:43:41 +02:00

.gitignore

initial commit

2026-09-27 11:20:34 +00:00

.oxlintrc.json

initial commit

2026-09-27 11:20:34 +00:00

index.html

initial commit

2026-09-27 11:20:34 +00:00

package-lock.json

initial commit

2026-09-27 11:20:34 +00:00

package.json

initial commit

2026-09-27 11:20:34 +00:00

README.md

half broken

2026-09-27 17:17:53 +00:00

tsconfig.app.json

initial commit

2026-09-27 11:20:34 +00:00

tsconfig.json

initial commit

2026-09-27 11:20:34 +00:00

tsconfig.node.json

initial commit

2026-09-27 11:20:34 +00:00

vite.config.ts

Fix console assets on nested routes

2026-09-27 22:51:04 +02:00

README.md

The console

Vite + React + TypeScript + Tailwind, built into internal/web/dist and served by the Go binary from an //go:embed. There is no separate deployment: the binary is the console.

Day-to-day

make web-check     # npx tsc -b --noEmit && npm run build  (what CI runs)
make web-dev       # Vite dev server, proxying /api to 127.0.0.1:8080
make web           # production build into internal/web/dist

For make web-dev to show anything, a fancy-dns has to be listening on 127.0.0.1:8080 with the API key the console sends. deploy/config.example.yaml binds there, but on port 53; rewrite the ports (or use make smoke, which does it for you) before starting it.

Two traps that have cost time, both from AGENTS.md:

tsc -b replays cached diagnostics from node_modules/.tmp/*.tsbuildinfo. After a large edit rm -f node_modules/.tmp/*.tsbuildinfo, or you will read errors that no longer exist and miss ones that do.
npm run build empties internal/web/dist and deletes its .gitkeep. Restore it (git checkout -- internal/web/dist/.gitkeep). The Go build still works while the directory is empty, so a console rebuild in the middle of someone else's go build is not fatal — but the commit is.

The API contract

src/lib/types.ts is a hand-written mirror of the Go json tags in internal/api and internal/app/hooks.go. apiFetch ends in JSON.parse(text) as T, so the compiler checks the console against its own declarations and nothing else: a field that loses its json tag, or a response a struct never had, shows up as an empty table in a browser and nowhere else. That is not hypothetical — a round of this found the Sources table, the ACL group cards and both Save buttons broken exactly that way.

So, when you touch either side:

Go: add or keep the json tag. internal/app has contract tests that assert the exact key set of every admin endpoint; they are the spec, and a rename fails them.
Console: change the interface in the same commit. If the server is wrong, fix the server — do not "adapt" in the client, and do not paper a mismatch over with as any, an optional field or a ?? 0 default. Those casts are precisely what hides drift.
Never rename a field on one side only. The dashboard, the ServiceMonitor queries and the Grafana panels are consumers of the same JSON.

The durable fix is to generate types.ts from the Go structs, or to publish an OpenAPI document from the server and generate from that. It is in the plan at the end of STATUS.md.

What is checked

npx tsc -b --noEmit (strict, with noUnusedLocals) and oxlint (two rules: react/rules-of-hooks and react/only-export-components). There is no test runner and no component test — the console's correctness rests entirely on its types, which are only as good as the mirror above. Adding vitest and a handful of tests for lib/format.ts (the percent/latency/duration formatters, which have already shipped a units bug) is cheap and worth doing.