[Shlink] `update` wipes shlink-web-client/servers.json, breaking the Web Client every time

[frontend-report]

Script: Shlink

Reported by: @Corgei

Guidelines

• Read and understood guidelines: yes
• Verbose mode used: yes

Issue occurs during: Script Update

Environment

• Linux distribution: Debian
• Proxmox version: 9.1
• Default settings: No
• Advanced settings: Yes
• Exact command: bash -c "$(curl -fsSL https://raw.githubusercontent.com/community-scripts/ProxmoxVE/main/ct/shlink.sh)"

Description

After running the built-in update command inside a Shlink LXC (installed via this repo's ct/shlink.sh / install/shlink-install.sh), the Shlink Web Client loses its configured server and shows the "Welcome! Add a server" screen, even though the Shlink API backend continues to work fine.

Root cause (confirmed by reading install/shlink-install.sh and ct/shlink.sh):

• On initial install, install/shlink-install.sh creates /opt/shlink-web-client/servers.json with the initial API key so the Web Client works out of the box.
• On update, update_script() in ct/shlink.sh does:
  1. For /opt/shlink (API): backs up .env and data/, does CLEAN_INSTALL=1 fetch_and_deploy_gh_release ..., then restores the backup. This correctly preserves configuration.
  2. For /opt/shlink-web-client (Web Client): does CLEAN_INSTALL=1 fetch_and_deploy_gh_release ... with NO backup/restore step for servers.json.
• Because CLEAN_INSTALL=1 wipes the target directory before redeploying, servers.json (and any other user customization under /opt/shlink-web-client, e.g. a custom nginx config placed there) is permanently deleted on every update. The Web Client then has no server configured and shows the onboarding screen, which looks like a "connection lost" / outage to the end user, even though nothing about the API, database, or network changed.

This is inconsistent with how /opt/shlink itself is handled (backup + restore around the clean install) and is easy to reproduce on every update.

Steps to reproduce

1. Install Shlink via this repo's script, opting in to "Install Shlink Web Client?" during setup.
2. Confirm the Web Client works and has a server configured (/opt/shlink-web-client/servers.json exists and is populated).
3. Run update from the container console (or re-run the ct script with the update path).
4. After the update finishes, open the Web Client in a browser.

Error output

No error/exception is thrown by the update script itself; it reports success. The symptom is only visible in the browser: the Web Client shows the onboarding screen ("Welcome! This application will help you manage your Shlink servers." / "+ Add a server") instead of the previously configured server, because /opt/shlink-web-client/servers.json no longer exists.

Verification performed on an affected container:
$ ls -la /opt/shlink-web-client/servers.json
ls: cannot access '/opt/shlink-web-client/servers.json': No such file or directory
$ curl -s http://127.0.0.1:3000/servers.json
(404 / empty — served via try_files /servers.json /conf.d/servers.json;, neither path exists after the update)

Meanwhile the API itself remains healthy:
$ curl -s -o /dev/null -w '%{http_code}\n' http://127.0.0.1:8080/rest/health
200

Additional context

Suggested fix, mirroring what is already done for /opt/shlink:

In ct/shlink.sh update_script(), before the CLEAN_INSTALL for shlink-web-client, back up /opt/shlink-web-client/servers.json (if it exists) to a temp location, and restore it after fetch_and_deploy_gh_release completes — the same backup/restore pattern already used for /opt/shlink/.env and /opt/shlink/data. For example:

if [[ -f /opt/shlink-web-client/servers.json ]]; then
cp /opt/shlink-web-client/servers.json /opt/shlink-web-client-servers.json.bak
fi
CLEAN_INSTALL=1 fetch_and_deploy_gh_release "shlink-web-client"
"shlinkio/shlink-web-client" "prebuild" "latest" "/opt/shlink-web-client"
"shlink-web-client_*_dist.zip"
if [[ -f /opt/shlink-web-client-servers.json.bak ]]; then
mv /opt/shlink-web-client-servers.json.bak /opt/shlink-web-client/servers.json
fi

As a workaround, I'm currently keeping a persistent copy of servers.json outside /opt/shlink-web-client and using a systemd path unit to restore it automatically whenever the directory changes, but this should ideally be fixed in the upstream update script so every user doesn't need to do this themselves.

[MickLesk]

To be honest, I don’t understand issues like this. You’ve probably spent a good 20 times as long describing this as you would have done simply adding the servers.json file after create_backup, haven’t you?

[Corgei]

To be honest, I don’t understand issues like this. You’ve probably spent a good 20 times as long describing this as you would have done simply adding the servers.json file after create_backup, haven’t you?

Nope, I thought it would be great if AI can help me do a bug report to you guys for explanations and findings so that you can fix / rectify / modify things easily. And that's all I could give you.

[MickLesk]

And why Not Just an Fork, edit and create PR? Thats faster then the AI debug

[Corgei]

And why Not Just an Fork, edit and create PR? Thats faster then the AI debug

Perhaps I'm not Developer and Maintainer of Proxmox VE Helper Scripts. Most importantly, I was thinking of not reporting anything in here since I didn't have enough time for my life but I still care for this project, if you don't want it, you can close it.

[Corgei]

@MickLesk I misclicked to an option for "Self-review completed – Code follows project standards." and GitHub bot closed the fix since I was not following the PR template lol.

[MickLesk]

Yeah bot is Sometimes stupid, i already reopen it