Open copumpkin opened 10 years ago
Sounds interesting indeed, although I don't yet see a nice way of declaring this. Please posts your ideas once they are fleshed out :)
Yeah, I have almost no experience with nixops yet, but in the abstract it seems like a general notion "concurrency level" within a certain scope seems like it would cover many of these cases. Will definitely post more once I gain more experience with the specifics :smile:
https://github.com/NixOS/nixops/issues/377 - looks like a clone of this issue
Note that this is now somewhat supported with https://github.com/NixOS/nixops/commit/f3f91c3f544dd1faaad1b724d42bc99dbb5aa3bb
An important part of a rolling restart feature is defining some health check condition and only progressing once health check succeeds.
Sometimes (rarely) I want a limited amount of control over the sequencing of operations NixOps performs. I might have a live database server that I want to make mostly declarative changes to, in a way that prevents all servers from being down at the same time. The obvious way to do this is to revert to imperative instructions, but I was thinking perhaps we could provide some way to specify declarative constraints of the sort like "only affect one server in this collection at a time". I haven't entirely fleshed out the idea, but it seems like a nice-to-have feature if someone can think of a way to make it work.