Restoring from a backup
Operating it
The steps to put a deployment back, read on the day it is needed.
Restoring is done from your own machine, never from the portal: a running deployment cannot replace its own database. You need nyxusctl, which an owner or administrator downloads under Restoring on Deployment › Backups; Docker, logged in to registry.nyxus.co.uk; and kubectl pointed at the cluster, or Docker Compose on the host.
Download the backup from your bucket or container.
Check that it opens. This changes nothing:
Restore it:
nyxusctl finds the deployment, shows every command it will run and asks before running them. --dry-run shows the plan and runs nothing.
What restore does:
The key stays on your machine. The backup is opened and checked there, in the receiver's own image with no network, and only the checked files are sent on.
On Kubernetes it stops the receiver, starts a short-lived pod on the receiver's volume, streams the files into it, puts them in place, removes the pod and starts the receiver again.
On Docker Compose it stops the receiver, puts the files in place with the service's own volume and user, and starts it again.
Nothing is deleted. The files a restore replaces are moved to
pre-restore-<time>/on the volume. Remove that folder once you are satisfied.If a step fails after the receiver has stopped,
restorestarts it again and says whether anything was replaced. If starting it again fails too, it says the receiver may still be stopped and shows the command that starts it.
While the receiver is stopped, collectors keep what they find and send it again, and the portal shows errors until it is back.
Nyxus upgrades a database when it starts, so a backup restores into the version that made it or any later one. It does not restore into an older version.
Into a new cluster
Install Nyxus at the same version or later, with the same values, the same secrets and the same key Secret, and let it start. Then run nyxusctl restore against it.