[bug]: Port 20426 is not available after 30 seconds.

[weberk]

No existing issues.

• There is no existing issue for my problem.

Describe the bug

dev-server watch is automatically terminating after first access to port 8081

reason is that hidden admin port 20426 has no process connected to.

To Reproduce

1. execute dev-server watch
2. verifiy there is a timeout connecting to port 20426
3. use the browser to access http://localhost:8081

Expected behavior

The admin GUI should be loaded.
Please advise how to return to a working version again. No backup available.

Screenshots & Logfiles

Port 20426 is not available after 30 seconds.
Starting browser-sync
[HPM] Proxy created: /browser-sync/**  -> http://127.0.0.1:22426
[HPM] Proxy created: /  -> http://127.0.0.1:20426
Starting web server on port 8081
...
2025-01-04 18:27:59.736  - info: host.dev-uvr16xx_bl-net-macBookPro.local received SIGINT

node:internal/process/promises:394
    triggerUncaughtException(err, true /* fromPromise */);
...
AxiosError: connect ECONNREFUSED 127.0.0.1:20426
    at AxiosError.from (/opt/homebrew/lib/node_modules/@iobroker/dev-server/node_modules/axios/dist/node/axios.cjs:876:14)
    at RedirectableRequest.handleRequestError (/opt/homebrew/lib/node_modules/@iobroker/dev-server/node_modules/axios/dist/node/axios.cjs:3156:25)
    at RedirectableRequest.emit (node:events:507:28)
    at eventHandlers.<computed> (/opt/homebrew/lib/node_modules/@iobroker/dev-server/node_modules/follow-redirects/index.js:49:24)
    at ClientRequest.emit (node:events:507:28)
    at emitErrorEvent (node:_http_client:104:11)
    at Socket.socketErrorListener (node:_http_client:512:5)
    at Socket.emit (node:events:507:28)
    at emitErrorNT (node:internal/streams/destroy:170:8)
    at emitErrorCloseNT (node:internal/streams/destroy:129:3)
    at Axios.request (/opt/homebrew/lib/node_modules/@iobroker/dev-server/node_modules/axios/dist/node/axios.cjs:4287:41)
    at process.processTicksAndRejections (node:internal/process/task_queues:105:5)
    at async /opt/homebrew/lib/node_modules/@iobroker/dev-server/dist/index.js:619:30 {
  port: 20426,
  address: '127.0.0.1',
  syscall: 'connect',
  code: 'ECONNREFUSED',
...
2025-01-04 18:28:00.349  - info: host.dev-uvr16xx_bl-net-macBookPro.local terminated

Adapter version

7.4.10

js-controller version

7.0.6

Node version

23.2.0

Operating system

MacOs Sequoia

Additional context

happend after accepting the update to admin adapter version 7.4.10 with the admin GUI
next dev-server watch was producing the described error situation
tried dev-server update, with no improvement

[mcm1957]

This (most likely) is NO bug of admin but related to dev-server. So it could be logged at https://github.com/[ioBroker/dev-server](https://github.com/ioBroker/dev-server). The proxy is an extension managed by dev-server and not in use during normal admin usage at target systems.

But in general I would reommend to discuss such questions at auf telegramm / discord channel first. Invote links can be found at www.iobroker.dev.

In addtion I would not expect that an gui based admin update within a running dev-server session is working. But thats also somethi g to dicuss at telegramm / discord.

[oweitman]

my observation showed that closing the open browser tabs used in a previous run for the dev server worked around the problem.

[weberk]

Closing the tab and even using a different browser (Chrome and Safari) did not changed the effect: dev-server stops with the mentioned exception on first access of localhost:8081.

I would say the admin.0 instance does not succeed to start?

await this.updateObject('system.adapter.admin.0', (admin) => {
           admin.native.port = this.getPort(adminPort, HIDDEN_ADMIN_PORT_OFFSET);
           admin.native.bind = '127.0.0.1';
           return admin;
       });

[weberk]

Workarround
Ich habs repariert, indem ich das .dev-server-Verzeichnis im Adapter-Basisverzeichnis gelöscht hab und den dev-server in dem Adapter neu konfigurieren hab lassen mit:
dev-server setup
Danach startete mein Adapter inklusive der admin.0-Instanz wieder gemeinsam und fehlerfrei.