| Commit message (Collapse) | Author |
|
|
|
They were accidentally added at the wrong location in dd4b6bcfc16c7c795b697195eb6703966352d9f4
|
|
These are the defaults, but I think it is better to explicitly state
them to ensure that we don't suffer from a mistake, when we think about
changing them in the future.
|
|
|
|
|
|
This is a manual fix until we get this merged into nixpkgs
Co-authored-by: Benedikt Peetz <benedikt.peetz@b-peetz.de>
|
|
Fixes: #87
|
|
|
|
instance list
This is necessary because of a bug in our instance which causes search
to fail under certain circumstances.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This uses the mastodon package from nixos-unstable-small because
backporting of a security release failed and we can't afford to wait.
|
|
|
|
|
|
|
|
|
|
Considering that the db dump takes longer than an hour, an hourly
service start time could lead to multiple dumps happening concurrently.
This should reduce this risk
|
|
Including this dump should remove the risk of a backup with a corrupt
PostgreSQL database. Initial test showed that the backup takes
around 32 GB and runs in under 3 hours.
There is one big oversight not yet included in this commit:
All services running a SQLite database are not included in this dump and
thus can not be safely recovered.
At present these are:
- etebase-server (db.sqlite3)
- murmur (murmur.sqlite)
(This is list was generated with `sudo fd sqlite /srv/`)
|
|
|
|
|
|
|
|
This isn't strictly necessary as we define the default config
|
|
nixpkgs
|
|
|
|
This doesn't work as nginx doesn't have the right permissions.
|
|
|
|
|
|
|
|
|
|
This is done to comply with the naming scheme employed at `vhack.eu`.
|
|
Otherwise, etebase might use the ipv6 ip, whilst nginx uses the ipv4
version. This prevents this issue
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This fully removes the human-factor and allows it to just run
`./generate` to generate all required certificates and keys (with the
needed extra keys and certificates)
|
|
Although the page does not actually serve any content, many browsers
will still refuse to access it at all, if they have the 'https-only'
mode activated.
|
|
|
|
|
|
|
|
|
|
|