Login page - Input fields refresh/empties after 5 seconds

[Sandalas]

Infos:

• Used Zammad version: 5.1.1-1651218076.e3861908.centos7
• Used Zammad installation type: yum
• Operating system: CentOS 7
• Browser + version:
  • Firefox 91.3.0esr
  • Chrome 90.0.4430.85
  • Edge 101.0.1210.32

Expected behavior:

• Login page won’t reload some of it’s DOM-Elements each 5 seconds

Actual behavior:

• The DOM element #app reloads its child elements every 5 seconds, this results in clearing the username and password fields
• if you are fast enough in typing, you can login

Steps to reproduce the behavior:

• Unknown

Background

We currently have several instances of Zammad running (as a VM).
One of these instances was unfortunately neglected, so it is still running on version 2.6.0.
I wanted to upgrade it to the latest version, but since it is a production instance, I wanted to test it with a clone first.
The update was done step by step, 2.6.0 → 3.0.0 → 3.6.0 → 4.0.0 → 5.1.1 considering all dependencies of Elasticsearch and Postgresql.
I have checked each version for functionality and the behavior only occurs since the update to 5.1.1.

The same problem existed here before, but was closed due to non-response → 2.9.x login issue after upgrade · Issue #2523 · zammad/zammad · GitHub

Answer to the asked questions in the old issue:

• Something in the productive.log: unfortunately no
  • Also nothing in zammad.access.log, zammad.error.log, nginx.log
• Are there change on files (e.q. CSS):
  • in /opt/zammad/ no
  • becaus it is a clone, I hat to change the hostname to -klon, and also changed it in the /etc/nginx/nginx.conf and /etc/nginx/conf.d/zammad.conf
  • created an self-signed certificate for the new hostname
• zammad run rake db:migrate was execuded
• zammad run rake searchindex:rebuild was execuded
• zammad run rake assets:precompile was execuded

[MrGeneration]

Sorry but that smells like a browser or addon specific issue.
I’ve never seen a Zammad instance do such a sketchy thing.

Using either incognito mode or safe mode of your browser could solve the issue.
If it does, disable your addons one by one to find the evil boii.

[Sandalas]

Thank you for your reply. This is exactly the kind of answer I was afraid of.

Currently I have 3 systems with the same version in use and only with this one system this problem occurs.

I have tested it with the 3 browsers mentioned above and now also again to be completely sure with a freshly installed Firefox without add-ons in SafeMode, but still the same problem.

---

Just now also once again performed a full update:

• kernel.x86_64 3.10.0-1160.66.1.el7
• zammad.x86_64 5.1.1-1655712708.d1d67c32.centos7
• elasticsearch.x86_64 7.17.4-1
• nginx.x86_64 1.20.1-9.el7
• postgresql96.x86_64 9.6.24-1PGDG.rhel7

---

Using the Firefox debugger I was able to trace the problem back to the “websocket” → “message” operation (see gif)
As soon as you jump into the function the input disappears.

Unfortunately I see nothing in the logs on the server nor in the console of the browser.

I hope this further information helps. If there is anything else you want me to test, let me know.

(This problem is so strange - I’m still hoping that this is just happening because it’s a cloned system with subsequently changed names).

[MrGeneration]

I honestly have no tips I can share on how to overcome this issue.
If this is a vanilla installation then the behavior indeed is strange.

Websocket messages shouldn’t cause a silent reload, usually this would happen (without silence) on a configuration change. However, this will trigger a modal. Possibly you have old cached files in your browser which may be your issue.

In case you wonder what modal I’m talking about, the maintenance section “Message” would show such a modal Maintenance — Zammad documentation