Hello,

I've reverted the Nixpacks version in use and the latest deploy looks to have went out successfully.

Side note, you do not need the `sleep 3` in your start command anymore.

Best,
Brody

Thanks - it was fixed on the revert. Is there some more info about what happened? Was there an update to nixpacks (and was there any announcement of this I missed re testing). Just want to understand if it's a one-off or if I should be doing something different in the future (or generally, what can be done to avoid the downtime).

Also - I was a bit surprised that this took down my app. Typical behavior is that until the new deploy is up and in some ready state, the old deployment stays alive and handles traffic, but a user reached out because the service went down.

Separately - while I'm thinking about it - is there a way to replace the default railway-branded page when the service isn't available?

Thanks,

R

Actually - when I try to deploy a similar but different service, I get the same error today (despite the original problem from yesterday being fixed - I haven't tried to redeploy that, don't want to crash it).

From what I can tell, the nixpacks configs are the same between the two. Switching to the "new builder" seems to have side stepped the issue though.

> Is there some more info about what happened? Was there an update to nixpacks

There was upstream updates in nix and when we updated nixpacks, whatever nix updated consequently broke your build, it is near impossible for us to test for these things, so sometimes some builds can break, we know this isn't acceptable and are currently working on an improved build pack.

> was there any announcement of this I missed

No announcement, we don't announce updates to Nixpacks because for 99.99% of users nothing happens, but thats not good enough for us hence the work on the new and improved build pack.

> Just want to understand if it's a one-off or if I should be doing something different in the future

It can happen again, for ultimate control over your build I would recommend looking into a Dockerfile, that way upstream changes to Nix that are out of both our and your control do not affect your build.

> I was a bit surprised that this took down my app.

That is totally understandable because you have yet to set a health check, without a health check we have no way to make sure your application is ready to serve traffic.

> is there a way to replace the default railway-branded page when the service isn't available?

Unfortunately no.

> when I try to deploy a similar but different service, I get the same error today (despite the original problem from yesterday being fixed - I haven't tried to redeploy that, don't want to crash it).

Makes sense a similar service would fail, I only rolled back the Nixpacks version on the service you linked, you can do this yourself with a NIXPACKS_VERSION=1.29.1 service variable, or of course take your build into your own hands with a Dockerfile!

> From what I can tell, the nixpacks configs are the same between the two.

Yeah, same plan, but different Nixpacks versions, the newer version is using using a different Nix archive and thats what is giving your build issues.

> Switching to the "new builder" seems to have side stepped the issue though.

This may sounds counter intuitive given you just said it stopped the issue, but we cannot recommend you use the New Builder right now, in fact, I though we removed that UI feature, the new builder is slower right now, and has many more issues that are not worth the hassle for yourself.

Great - thanks - will look into dockerfile, that seems the slightly more complicated but correct solution.

Sounds good!

Please re-open if you have any more questions!