I notice @anon55243134 has deleted almost all their posts. I really think there are lessons to be learned here for the team and for maintenance of the update scripts and the messaging around updating.
@anon55243134 is someone who has been running a self-hosted discourse for years and has now got a damaged and non-functioning installation - just by following the prompts to upgrade.
If that happened to me I’d be very annoyed and distressed at potentially losing my forum contents. Having opted for self-hosting I might not be ready or able to pay big money to get it fixed, if that’s even possible.
I think there are insufficient warnings and checks
- has the user taken a recent backup (not a hosting services snapshot!)
- has the user downloaded it
- is the user told that the web-based update might fail and require a command line update
- is the user asked to check if their OS is very old
- is the user told that migrating to a new up to date server might end up being the best approach
- is the user warned that major updates (such as a database update) can be perilous and if inexperienced waiting a week might be a good idea, for problems to be found and fixed
More worrying still, in one of the deleted posts I see some pretty dramatic failures which were not trapped and the script continued:
cat: /shared/postgres_data/PG_VERSION: No such file or directory
...
E: Unable to locate package postgresql--pgvector
cp: cannot stat '/etc/postgresql//main/*': No such file or directory
sh: 1: /usr/lib/postgresql/bin/postgres: not found
...
Finding the real data directory for the source cluster
could not get data directory using "/usr/lib/postgresql/bin/postgres" -D "/shared/postgres_data" -C data_directory: No such file or directory
Failure, exiting
I haven’t checked the scripts, but I would expect things not existing is an indication that trouble lies ahead, and it’s time to stop.