Nextcloud update pushed too early?

[firmansi]

So far I have no issues as mentioned above with latest update, my stack running with onlyoffice, nextcloud assistant, team folder, draw.io, groupware. My NC running with more than 100 users with Cloudron

[avatar1024]

I do see this show up several times in the log when that nextcloud instance starts:

Healtheck error: Error: connect ECONNREFUSED 172.18.18.24:80

[nebulon]

@avatar1024 that is to be expected. This is the Cloudron platform trying to reach the app to check if it is already up and running. When you see this, it just means the app isn't responding just yet.

[avatar1024]

@nebulon Is it somehow possible to reinstall the app with the same version (or rerun the update but without starting from an old backup)?

[avatar1024]

Right so it's fixed on that broken. It turns out that for some reasons the ownership on many directories inside the NC data directory have been changed from www-data to yellowtent. This included the content of: data/appdata_{instanceid}/appstore/ which looked like:

drwxr-xr-x  3 yellowtent yellowtent 4.0K Oct  2 12:55 .
drwxr-xr-x 19 www-data   www-data   4.0K Oct  2 12:55 ..
-rw-r--r--  1 yellowtent yellowtent  87K Oct  1 11:19 appapi_apps.json
drwxr-xr-x  5 yellowtent yellowtent 4.0K Oct  1 11:19 app-discover-cache
-rw-r--r--  1 yellowtent yellowtent 2.8M Oct  1 11:19 apps.json
-rw-r--r--  1 yellowtent yellowtent 161K Oct  1 11:19 categories.json
-rw-r--r--  1 yellowtent yellowtent 7.2K Oct  1 11:19 discover.json

...and which broke the App Store. Revering the permission to www-data made everything work right away.

But actually many directories are affected inside the appdata_{instanceid} directory as well as users files inside the data directory.

I'm going to do a good backup and put the app in maintenance mode before I change ownership every inside the data because it feels a bit scary.

@staff Any clues on how it happened??

[avatar1024]

@staff quick question. When Nextcloud 32.0.1 gets released, will it be possible to update directly to it for those instance that are still on NC31, skipping 32.0.0?

[girish]

@avatar1024 good idea, we can revoke 32.0.0 so it can be skipped.

[avatar1024]

@girish on my instances NC still tries to first update to 32.0.0 despite 32.0.1 having now been pushed.

[nebulon]

@avatar1024 thanks for the reminder. Since the patch release package is out now, I've revoked the previous one. The update should now directly hop onto 32.0.1 (maybe check manually for updates first via the dashboard)

[ekevu123]

Upgrading to 32.0.1 worked well for me.