# Multiple issues with upgdates came up while I have been developing an app to handle this. — retained record

> Retained reddit record from reddit:selfhosted.

Canonical: https://highsignal.app/data/records/5cb0ed22cfe34ddf

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:selfhosted)  published 2026-10-24 21:15 ingested 2026-10-25 02:37

[original source ↗](https://www.reddit.com/r/selfhosted/comments/1xek3lg/multiple_issues_with_upgdates_came_up_while_i/)

## retained text



```
Im developing an app, that is admittedly going to be closed source, and iphone only, and cost some money.  However one of the key things I want it to do is to make updates for users safer and still be just a couple taps.  This has exposed issues I have with my own manual updates as well as issues I have seen since I am supporting some services I don't run myself.  I figured that others may be able to benefit from my headaches.

  
So original I was just going to make the app auto pull new images, which if you have been selfhosting for more than a minute you should know was a terrible first idea.  Once I made it more perscriptive I still had issues.

Jellyfin - Now I am a plex user myself, but for an app I don't want to force people to use plex, so jellyfin.  Jellyfin it turns out answers to a /health with 200 (passed/good) even when its still in a migration or in start up.  This can cause failed installs and upgrades to not be caught by automation.  I had to switch the endpoint I was checking in order to not have this as an issue

Nextcloud - I know everyone is suprised I had issues with upgrading nextcloud :-).  The biggest issue was that nextcloud can't jump over major upgrades.. you can't go from 33 - 35.  So I had to check for the upgrades in between, backup the database due to scheme changes, do the update, test that step, if passed backup again and do the next, until I reached most current at which point I could clear out all the backups I didn't need.  With out that check each step I couldn't roll back and I ruined more test installs than I care to admit (4).

I can't just upgrade databases between versions.  Postgress is at version 18 but I was doing everything on 16.  Not every service has been tested against or works with 18.  Since I test every service against 9 different OS/CPU combinations this caused a huge headache as I had to have different tests and pinned versions for different services.  I also couldn't have any services share a database because that could potentionally break other services when just updating one.

You can't undo every upgrade.  This is Nextcloud again.  Once you let the database migrate settings up to the newer versions, you can't just downgrade without the full save from that version, which I still clean out even knowing this, just be aware.

If your drive is full it can look like a hang to automation.  Why isn't that test finishing.. troubleshoot the script.. find nothing wrong.. still won't finish.. oh I filled the whole vm's drive with docker images and can't fit another.  I solved this myself with snapshots in proxmox of my test systems to just make sure they are 'clean' everytime.

Dockers pull limit is real.  100 pulls an hour.. when your tests pull 18 images, and do so for 9 different vms you hit that limit and have to put in controled spacing of an hour to not have that be an issue.

If you have any upgrade or roll back pitfalls or warnings let me know.  Thanks
```



document fetched 2026-10-25 00:27 · canonical https://reddit.com/r/selfhosted/comments/1xek3lg/multiple_issues_with_upgdates_came_up_while_i
