Upgrade fails when reloading nginx with "invalid port in upstream"

[MayeulC]

2022-11-17 10:54:14,716: DEBUG - nov. 17 10:54:14 nginx[1553622]: nginx: [emerg] invalid port in upstream "127.0.0.1:/monitorix" in /etc/nginx/conf.d/maindomain.tld.d/monitorix.conf:2

https://paste.yunohost.org/raw/erulayaxec

More recent attempt: https://paste.yunohost.org/raw/ufahozocuk

_Originally posted by @MayeulC in https://github.com/YunoHost-Apps/monitorix_ynh/issues/40#issuecomment-1318383205_

[MayeulC]

Still an issue: https://paste.yunohost.org/raw/hahofuxima

yunohost_version: 11.0.11
  YNH_APP_CURRENT_VERSION: 3.12.0~ynh2
  YNH_APP_MANIFEST_VERSION: 3.15.0~ynh2

Edit: same with yunohost 11.1

[MayeulC]

It failed restoring too: https://paste.yunohost.org/raw/piyupoqule

Looking at the log, //tmp/tmp.0CDrYHYynC/app.deb is obviously invalid, and only /tmp/tmp.0CDrYHYynC/app exists.

Edit: fixed the restore issue by manually editing the last line of that part of the restore script:

tempdir="$(mktemp -d)"
ynh_setup_source --dest_dir="$tempdir"
ynh_package_update
dpkg --force-confdef --force-confold -i /$tempdir/app.deb

to:

dpkg --force-confdef --force-confold -i $tempdir/app

[MayeulC]

SOLVED!

I only had the port: 8081 yunohost setting; the http_port setting was not set.

I suggest:

• setting set -u to disallow empty variables in bash. This may have side effects in the scripts, luckily the CI should catch these, and it should be easy to check. Maybe even set -euox pipefail
• in the upgrade script, migrate from the old "port" to the new "http_port" yunohost setting variable

Can I delete the old "port" setting key?

[Josue-T]

Closing as theses settings was mainly reworked with packaging v2 and #55