@tdk-landscape/tdk-cli-core 1.3.61 → 1.3.62

package/CHANGELOG.md

@@ -4,6 +4,10 @@ Notable changes to the `tdk` CLI. Versions match [npm](https://www.npmjs.com/pac
 
 ## Unreleased
 
+- `tdk doctor` no longer skips the service health check when Traefik answers with an error. A 404 (no route yet) or 502/503/504 (route found, container not answering) is reported as a failure that names the service and URL, instead of being read as "Traefik never bound :80". It is skipped only when every probe fails to connect.
+
+## 1.3.61 (2026-09-28)
+
 - `tdk doctor` and `tdk up` no longer report free host ports (80, 443, 5432) as "taken" on Linux. The port check bound `0.0.0.0` and `127.0.0.1` at the same time, and on Linux the second bind collided with our own first probe. It now probes them one after the other, which still catches a loopback-only Postgres on macOS. Found by the nightly clean-machine quickstart run.
 
 ## 1.3.60 (2026-09-27)

package/cli/dist/commands/doctor.js

@@ -4,13 +4,13 @@ import { dirname, join, normalize } from "node:path";
 import chalk from "chalk";
 import { Command } from "commander";
 import { MASTER_CONFIG_FILES, QUICKSTART_DOCS_URL, REQUIRED_PACKAGE_SCRIPTS, } from "../utils/constants.js";
-import { checkHostPorts, checkIngressPorts, checkPrivateNpmRegistry, checkTiltResourceHealth, } from "../utils/doctor-runtime.js";
+import { checkHostPorts, checkIngressPorts, checkPrivateNpmRegistry, checkTiltResourceHealth, summarizeServiceProbes, } from "../utils/doctor-runtime.js";
 import { validateEnvFile } from "../utils/env-validator.js";
 import { formatCount } from "../utils/formatting.js";
 import { findProjectRoot } from "../utils/paths.js";
 import { buildHealthTargets, pingHealthTargets } from "../utils/service-urls.js";
 import { discoverResourcesFromRoot } from "../utils/services.js";
-export { checkIngressPorts, checkPrivateNpmRegistry, checkTiltResourceHealth, summarizeTiltBuildError, } from "../utils/doctor-runtime.js";
+export { checkIngressPorts, checkPrivateNpmRegistry, checkTiltResourceHealth, summarizeServiceProbes, summarizeTiltBuildError, } from "../utils/doctor-runtime.js";
 // A wedged Docker daemon makes `docker ps` block forever instead of failing,
 // and doctor is exactly the tool people run when their environment is broken.
 const EXEC_TIMEOUT_MS = 10_000;
@@ -658,41 +658,7 @@ async function checkServiceHealth(timeoutMs) {
             message: "No routable services to ping",
         };
     }
-    const probes = await pingHealthTargets(targets, timeoutMs);
-    const reachable = probes.filter((probe) => probe.ok);
-    const failed = probes.filter((probe) => !probe.ok);
-    // Nothing answered at all: either the stack was never started, or ingress
-    // never came up (Traefik port conflict). Prefer failing via
-    // checkTiltResourceHealth / checkIngressPorts when Tilt is up; here we only
-    // skip when there is truly nothing to probe yet.
-    if (reachable.length === 0) {
-        return {
-            name: "Service Health",
-            didPass: true,
-            isSkipped: true,
-            message: `No services responded - skipped ping of ${formatCount(targets.length, "service")} (stack may be down, or Traefik never bound :80)`,
-            fix: "If `tdk up` is already running, check Traefik/port 80 in the Tilt UI. Otherwise start with: tdk up",
-        };
-    }
-    if (failed.length === 0) {
-        return {
-            name: "Service Health",
-            didPass: true,
-            message: `All ${formatCount(targets.length, "service")} responding on /health`,
-        };
-    }
-    const details = failed
-        .map((probe) => {
-        const reason = probe.status ? `HTTP ${probe.status}` : (probe.error ?? "no response");
-        return `${probe.name} (${reason})\n      ${probe.url}`;
-    })
-        .join("\n    ");
-    return {
-        name: "Service Health",
-        didPass: false,
-        message: `${formatCount(failed.length, "service")} not responding (${reachable.length}/${probes.length} healthy):\n    ${details}`,
-        fix: "Check container state and routing: docker ps, then tdk networks to compare the advertised URLs against Traefik's routers",
-    };
+    return summarizeServiceProbes(await pingHealthTargets(targets, timeoutMs));
 }
 export const doctorCommand = new Command("doctor")
     .description("Check environment readiness for TDK")

package/cli/dist/utils/doctor-runtime.js

@@ -349,6 +349,61 @@ export async function probeHostPort(port) {
         return wildcard;
     return probeAddress(port, "127.0.0.1");
 }
+/** What an HTTP status from a service's health URL means, in the words a user needs. */
+function describeProbeFailure(probe) {
+    const { status } = probe;
+    if (status === undefined)
+        return probe.error ?? "no response";
+    if (status === 404) {
+        return "HTTP 404: Traefik is answering but has no route for this URL (container not started yet, or not running)";
+    }
+    if (status === 502 || status === 503 || status === 504) {
+        return `HTTP ${status}: Traefik found the route but the container isn't answering (still starting, or unhealthy)`;
+    }
+    return `HTTP ${status}`;
+}
+/**
+ * Turns health-URL probes into a doctor result.
+ *
+ * "Nothing answered" (connection refused or timeout on every probe) means the
+ * stack is down or Traefik never bound :80, so the check is skipped. Any HTTP
+ * response counts as an answer, including a 404 or 502 from Traefik: that means
+ * the ingress is up and the service behind it is not reachable, which is a
+ * failure. Treating only 2xx as "responded" reported a running Traefik with no
+ * route as "Traefik never bound :80" and skipped the check.
+ */
+export function summarizeServiceProbes(probes) {
+    const answered = probes.filter((probe) => probe.status !== undefined);
+    const failed = probes.filter((probe) => !probe.ok);
+    if (answered.length === 0) {
+        return {
+            name: "Service Health",
+            didPass: true,
+            isSkipped: true,
+            message: `No services responded - skipped ping of ${formatCount(probes.length, "service")} (stack may be down, or Traefik never bound :80)`,
+            fix: "If `tdk up` is already running, check Traefik/port 80 in the Tilt UI. Otherwise start with: tdk up",
+        };
+    }
+    if (failed.length === 0) {
+        return {
+            name: "Service Health",
+            didPass: true,
+            message: `All ${formatCount(probes.length, "service")} responding on /health`,
+        };
+    }
+    const details = failed
+        .map((probe) => `${probe.name} (${describeProbeFailure(probe)})\n      ${probe.url}`)
+        .join("\n    ");
+    const notRouted = failed.some((probe) => probe.status === 404);
+    return {
+        name: "Service Health",
+        didPass: false,
+        message: `${formatCount(failed.length, "service")} not healthy (${probes.length - failed.length}/${probes.length} healthy):\n    ${details}`,
+        fix: notRouted
+            ? "If `tdk up` just started, images may still be building: re-run `tdk doctor` in a minute. Otherwise check `docker ps` and the resource in the Tilt UI, and compare `tdk networks` with the URL above."
+            : "Check container state and routing: docker ps, then tdk networks to compare the advertised URLs against Traefik's routers",
+    };
+}
 /**
  * Catches host ports taken by something other than this project's containers,
  * most often a local Postgres on 5432 or a web server on 80. Docker-held 80/443

package/cli/package.json

@@ -1,7 +1,7 @@
 {
   "name": "tdk-cli-internal",
   "private": true,
-  "version": "1.3.61",
+  "version": "1.3.62",
   "description": "TDK CLI - Tilt Development Kit command line interface",
   "keywords": [
     "tdk",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tdk-landscape/tdk-cli-core",
-  "version": "1.3.61",
+  "version": "1.3.62",
   "description": "Run 100 microservices on a 16 GB laptop without Kubernetes. Scaffold services and run them locally with hot reload, built on Tilt.",
   "homepage": "https://tdk-landscape.github.io/tdk-website",
   "repository": {