From d17640e861f1d5ac11076d0568990a42aacb3e69 Mon Sep 17 00:00:00 2001
From: Doug Horner <horner@mieweb.com>
Date: Sat, 5 Sep 2026 23:03:03 -0400
Subject: [PATCH 1/2] fix(netbox): retry network failures and surface fetch
 error cause

undici reports DNS/connect/TLS failures as a bare 'fetch failed', hiding
the real reason in err.cause. Log the cause and URL, and retry
network-level failures (3 attempts, backoff). HTTP error responses are
not retried.

Fixes #469
---
 create-a-container/utils/netbox.js | 33 +++++++++++++++++++++++++++++-
 1 file changed, 32 insertions(+), 1 deletion(-)

diff --git a/create-a-container/utils/netbox.js b/create-a-container/utils/netbox.js
index 1661615f..c3820f72 100644
--- a/create-a-container/utils/netbox.js
+++ b/create-a-container/utils/netbox.js
@@ -21,6 +21,11 @@ const NETBOX_COMMENT = 'This container was built using opensource-server';
 // for display, so convert gigabytes to MB before sending.
 const MB_PER_GB = 1000;
 
+// Network-level failures (DNS blip, hairpin-NAT connect reset) are retried;
+// HTTP error responses are not.
+const FETCH_ATTEMPTS = 3;
+const FETCH_RETRY_DELAY_MS = 1000;
+
 /**
  * Build request headers for NetBox API calls.
  * @param {string} token - NetBox API token
@@ -33,6 +38,32 @@ function headers(token) {
   };
 }
 
+/**
+ * fetch() with retries on network errors. undici reports these as a bare
+ * "fetch failed", so the underlying cause is surfaced in the thrown message.
+ * @param {string} url
+ * @param {object} init - fetch init options
+ * @returns {Promise<Response>}
+ */
+async function fetchWithRetry(url, init) {
+  let lastErr;
+  for (let attempt = 1; attempt <= FETCH_ATTEMPTS; attempt++) {
+    try {
+      return await fetch(url, init);
+    } catch (err) {
+      lastErr = err;
+      if (attempt < FETCH_ATTEMPTS) {
+        await new Promise(r => setTimeout(r, FETCH_RETRY_DELAY_MS * attempt));
+      }
+    }
+  }
+  const cause = lastErr.cause?.code || lastErr.cause?.message || lastErr.message;
+  throw new Error(
+    `NetBox unreachable after ${FETCH_ATTEMPTS} attempts (${init.method || 'GET'} ${url}): ${cause}`,
+    { cause: lastErr },
+  );
+}
+
 /**
  * Perform a fetch against the NetBox API.
  * Throws on non-2xx responses. Returns null for 204 No Content.
@@ -44,7 +75,7 @@ function headers(token) {
  */
 async function nbFetch(baseUrl, token, path, options = {}) {
   const url = `${baseUrl.replace(/\/$/, '')}/api${path}`;
-  const res = await fetch(url, {
+  const res = await fetchWithRetry(url, {
     ...options,
     headers: { ...headers(token), ...(options.headers || {}) },
   });

From 6c7326e37661c09d213030c73eec990223c51906 Mon Sep 17 00:00:00 2001
From: Doug Horner <horner@mieweb.com>
Date: Sat, 5 Sep 2026 23:07:05 -0400
Subject: [PATCH 2/2] fix(netbox): only retry undici network errors

Rethrow non-network exceptions (bad URL, abort) immediately instead of
retrying them and masking the original error.
---
 create-a-container/utils/netbox.js | 15 ++++++++++++---
 1 file changed, 12 insertions(+), 3 deletions(-)

diff --git a/create-a-container/utils/netbox.js b/create-a-container/utils/netbox.js
index c3820f72..3f42b296 100644
--- a/create-a-container/utils/netbox.js
+++ b/create-a-container/utils/netbox.js
@@ -39,8 +39,16 @@ function headers(token) {
 }
 
 /**
- * fetch() with retries on network errors. undici reports these as a bare
- * "fetch failed", so the underlying cause is surfaced in the thrown message.
+ * undici wraps DNS/connect/TLS failures as TypeError('fetch failed') with the
+ * real reason in err.cause. Anything else (bad URL, aborted, etc.) is not
+ * a transient network error.
+ */
+function isNetworkError(err) {
+  return err instanceof TypeError && err.message === 'fetch failed' && err.cause != null;
+}
+
+/**
+ * fetch() with retries on network errors. Other errors are rethrown immediately.
  * @param {string} url
  * @param {object} init - fetch init options
  * @returns {Promise<Response>}
@@ -51,13 +59,14 @@ async function fetchWithRetry(url, init) {
     try {
       return await fetch(url, init);
     } catch (err) {
+      if (!isNetworkError(err)) throw err;
       lastErr = err;
       if (attempt < FETCH_ATTEMPTS) {
         await new Promise(r => setTimeout(r, FETCH_RETRY_DELAY_MS * attempt));
       }
     }
   }
-  const cause = lastErr.cause?.code || lastErr.cause?.message || lastErr.message;
+  const cause = lastErr.cause.code || lastErr.cause.message;
   throw new Error(
     `NetBox unreachable after ${FETCH_ATTEMPTS} attempts (${init.method || 'GET'} ${url}): ${cause}`,
     { cause: lastErr },