Nodes aren't started about docker-machine-driver-hetzner HOT 6 CLOSED

I've let cluster creation go on and the nodes that I manually started are shut down again. It appears Rancher is shutting them down for some reason.

from docker-machine-driver-hetzner.

Further insight reveals that the nodes are created without an interface to the 10.0.0.0/16 network, despite having selected that private network in the node template. Therefore Rancher has no way of even connecting to these nodes, but I still don't understand why Rancher shuts them down, after I start them manually.

from docker-machine-driver-hetzner.

Hi,

sorry for the inconvenience the driver is causing you. From the driver's perspective, machine creation, starting instances and shutting them down are distinct operations its RPC interface has to expose.

I'm unfamiliar with the Rancher side of things, but could you perhaps try to create a machine manually using docker-machine and setting appropriate flags to create the private network and use it for communications? Ideally from another cloud instance on the same network, as to allow the whole provisioning process to run (if that is feasible for you from the cost perspective). If that fails as well, perhaps something in the hcloud API changed or we introduced a regression some time ago that nobody noticed has noticed yet. On the other hand, if that works fine, something is going wrong down the line with ranger.

The driver binary itself is rather 'dumb' and does not do any fancy timeout stuff (or automatic shutdowns for that matter), so from the behavior you describe I can only guess that perhaps Rancher is giving up after some time and then calls the shutdown method on the driver to get rid of the servers -- although that would not really make sense, as Destroy() would be more appropriate here.

from docker-machine-driver-hetzner.

I don't think the driver is causing any problems. I thought so in the beginning, because I assumed the Hetzner Cloud would work like it always did. However, it appears that they actually introduced a bug a couple of days ago, which they are fixing right now. I'll report back once they fixed it.
Do you know of a way to watch the provisioning process via some logfile or other mechanism?

from docker-machine-driver-hetzner.

docker-machine has a --debug flag that can be used with the create subcommand. That should output both the log information from the driver and the provisioning process itself.

from docker-machine-driver-hetzner.

Thanks and I close this issue, because Hetzner fixed something and now I cannot reproduce the problem.

from docker-machine-driver-hetzner.