[Nix-dev] NixOps usage survey.

[zimbatm]

I've been using NixOPS for my personal servers (on Vultr) and I couldn't be
happier. I'm deploying a rails app and I never had a better experience than
that, even on Heroku. The initial setup take a bit longer but it already
paid back in terms of no-brainer deploys. I love it that I can share
configuration between my own machine and servers and not have to worry
about state.

In a work context I think I would need a couple of more things.

Because of how the nixos activation scripts are working it's quite hard to
do zero-downtime deploys unless you have a load-balancer in front. Right
now it takes down the old version and then boots the new version. It's
simple but it also leaves a gap where users get 503 status codes.

I would probably also setup a "deploy" gateway machine to keep the
deployment centralized.

One issue is that because NixOps is using a push model it doesn't work very
well with auto-scaling. Usually in that setup the machine is pulling the
configuration on boot. I think with a proper API interface on the deploy
gateway it should be possible to handle that case but it's not supported
out of the box.


On Wed, 7 Sep 2016 at 11:37 Игорь Пашев <pashev.igor at gmail.com> wrote:

> 2016-09-07 13:22 GMT+03:00 4levels <4levels at gmail.com>:
> > We've been pleasantly surprised that every single deploy has given
> identical
> > results on all (currently 12) servers in the various datacenter
> locations of
> > Vultr.
>
>
> Yeah, develop in VBox, deploy to everywhere. The other day I moved one
> of the machines from EC2 to Hetzner... in the blink of an eye :-)
> _______________________________________________
> nix-dev mailing list
> nix-dev at lists.science.uu.nl
> http://lists.science.uu.nl/mailman/listinfo/nix-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.science.uu.nl/pipermail/nix-dev/attachments/20160907/fe352a40/attachment.html>