By default, Preview hosted deployment clusters run under the
envs.dev domain. Every .dev domain is on the HSTS preload list, which makes HTTPS required on all connections. Options to correct this are:
Recommended: Enable HTTPS on your application. This is as simple as adding the
kubernetes.io/tls-acme: "true" annotation on your Chart's Ingress.
Use a non-Preview cluster with a custom hostname. This will allow you to use a domain which is not on the HSTS preload list.
We recognize this is a problem, and are looking to get a better feedback loop here. Part of the issue here is that we attempt to clean up failed deployments, which can make debugging difficult.
In your config file, you can set
deploy.atomic: false, which will prevent automatic clean up of failed deployments. Note that this can leave deployments in a state that requires manual intervention in order to deploy again, so we do not recommend using this setting when not actively debugging.
If you are having difficulty debugging deployments, reach out and we'll jump on helping you figure out what went wrong.
Preview environments have resource limits in place. These limits may prevent a pod from scheduling it if exceeds them. At the moment these limits can not be customized, but please let us know if you are hitting them.