Connection refused from Cypress dashboard makes E2E job fail

[cnotv]

Description

E2E jobs seems all failing due connection refused. Example errors:

Cypress failed to make a connection to the Chrome DevTools Protocol after retrying for 50 seconds.

This usually indicates there was a problem opening the Chrome browser.

The CDP port requested was 44057.

Error: connect ECONNREFUSED 127.0.0.1:44057
    at TCPConnectWrap.afterConnect [as oncomplete] (node:net:1187:16)

  Running:  pages/about.spec.ts                                                            (1 of 14)
read ECONNRESET
Error: read ECONNRESET
    at TCP.onStreamRead (node:internal/stream_base_commons:217:20)

The issue is caused by SorryCypress server for some reasons, although the run are logged (see attachment).

The only change occurred in this time, based on E2E retry of my branch retry 1 and 2, seems to be a new release 4 days ago (see attachment).

[cnotv]

The issue may be related to some configuration changes, as it seems now it's required to be done in a different way, including the URL.

[cnotv]

Making an educated guess, the issue may be related to cy2 version not pinned anymore 🙃 It does not make any difference.

Another cause may be related to this reported bug.
Updating Cypress I got this

------------------------------------------------------------------------------
From the Cypress Team:

Beginning June 30, 2023, recording Cypress tests to self-hosted solutions now requires the Cypress Gateway Connector. 

To continue recording, please visit the link below:

https://on.cypress.io/gateway-connector?id=d2ZzaG1qdzJpZnhtZ3Rmd2k=

Not sure why it happens now. It seems like Cypress is trying to cut them out, so we'll have to configure what is written in the previous comment.

In a latest update, the issue is related to a Chrome version not working with Cypress <12.15.0 🥵

[cnotv]

To pin a specific version of Chrome is required to install locally or use provided Docker images.
This leaded to a sequence of different issues, mainly around /bin/sh: 1: docker: not found as reported in the job from the CI.

Given the current state caused by Chrome release 117.0.5851.0, I rather aim to disable the dashboard till they fix it and pin the browser version in another time.