| 13 Nov 2023 |
 raitobezarius | We need to write:
- documentation for migration
- send an announcement for unstable users
- release notes for 23.11
ASAP
| 13:44:28 |
| raitobezarius | I will do the modifications to Ma27's PR to go with this | 13:44:42 |
 bendlas | good call! | 13:47:10 |
| bendlas |
- let's please steer users away from
initialScript. Maybe even deprecate as well (remember, this is only run once for the whole cluster, so huge pitfall when adding new DBs)
| 13:49:18 |
| bendlas | maybe do something like repeatableInit instead | 13:49:53 |
| raitobezarius | In reply to @bendlas:matrix.org
- let's please steer users away from
initialScript. Maybe even deprecate as well (remember, this is only run once for the whole cluster, so huge pitfall when adding new DBs)
Please do send a PR to postgresql.md to explain how to do this | 13:51:31 |
| raitobezarius | I will cherry-pick your commit with Ma27 PR or we will figure out something | 13:51:41 |
| raitobezarius | I don't think we should deprecate initialScript right now | 13:51:55 |
| raitobezarius | All deprecations are expensive and we are running out of consideration budgets :P | 13:52:05 |
| raitobezarius | But let's write how to migrate from ensurePermissions to postStart in a palatable way | 13:52:17 |
| bendlas | Yep, I can write something up. postStart is also not ideal, since it only lets you talk about ordering using mkAfter and the like, but should be enough for most people for now. | 13:55:31 |
| bendlas | maybe we can get a few people together to rethink all of ensure* across postgresql and mysql (and others?) | 13:56:22 |
| bendlas | ... for 24.05 that is | 13:56:28 |
| raitobezarius | Yep I am moving all the thinking to 24.05 | 13:57:29 |
| raitobezarius | In reply to @bendlas:matrix.org
Yep, I can write something up. postStart is also not ideal, since it only lets you talk about ordering using mkAfter and the like, but should be enough for most people for now. I think so too | 13:57:39 |
| raitobezarius | Like it's not less expressive than ensurePermissions | 13:57:46 |
| bendlas | true | 13:58:11 |
| bendlas | ... except .. ensure* always puts permissions after db creation, but oh well ... | 14:00:00 |
| raitobezarius | you can always do the permission creation on the preStart of the module you are interested | 14:03:45 |
| raitobezarius | or you can have a postgresql-$myservice-install-permissions.service oneshot thingie | 14:03:56 |
| raitobezarius | I posted https://discourse.nixos.org/t/breaking-changes-announcement-for-unstable/17574/39?u=raitobezarius | 14:05:10 |
| bendlas | In reply to @raitobezarius:matrix.org
or you can have a postgresql-$myservice-install-permissions.service oneshot thingie good point. I'm wondering if we should advise people to do that instead of mkOrder sorcery | 14:06:26 |
| raitobezarius | Yep, I think that might be the best option | 14:06:43 |
| raitobezarius | We already have people doing local ensure logic in their own modules depending on the circumstances | 14:07:10 |
| raitobezarius | We just move more complexity to NixOS modules authors | 14:07:18 |
| bendlas | only fly in the ointment: With pre-15 psql, where db user isn't the owner, I think this wouldn't work, creating yet another lockstep - constraint on psql version :/ | 14:10:05 |
| raitobezarius | we only need to provide the information | 14:10:55 |
| raitobezarius | nixpkgs nixos modules are fine | 14:10:57 |
| raitobezarius | non-nixpkgs nixos modules will probably have issues | 14:11:05 |
| raitobezarius | we have to write documentation with this in mind | 14:11:09 |
