Wait on srv.NextActions after creating server

[apricote]

Julian from @hetznercloud here 👋

while investigating a customer ticket, I noticed that you are not waiting on the additional actions returned from the POST /v1/servers endpoint. Instead you poll the server to wait for it to startup.

docker-machine-driver-hetzner/driver/driver.go

Lines 500 to 511 in 54b321f

	log.Infof(" -> Creating server %s[%d] in %s[%d]", srv.Server.Name, srv.Server.ID, srv.Action.Command, srv.Action.ID)
	if err = d.waitForAction(srv.Action); err != nil {
	return errors.Wrap(err, "could not wait for action")
	}
	d.ServerID = srv.Server.ID
	log.Infof(" -> Server %s[%d]: Waiting to come up...", srv.Server.Name, srv.Server.ID)
	err = d.waitForRunningServer()
	if err != nil {
	return err
	}

I would recommend you wait on the actions in srv.NextActions. In there is also the start_server action, which is marked as completed once the server is running. You can check out our CLI code for comparison: https://github.com/hetznercloud/cli/blob/50fe4ca15deea739279539cf0bb2628ee2693f0e/internal/cmd/server/create.go#L100-L110

There is also a helper client.Action.WatchOverallProgress that can watch multiple actions at the same time.

Based on the contents, the GET /v1/servers/:id endpoint which you currently use is less efficient than simply polling the actions. It also counts towards a different Rate Limit budget, which caused an issue for one customer because they exhausted the normal Rate Limit with these calls.

[JonasProgrammer]

Hi Julian,

thanks a lot for the input, I'll look into fixing this. Unfortunately, the way this currently works is yet another left-over from pre-hcloud-go times...

While I will provide a bug-fix release for this issue, I now seriously do consider re-implementing the whole creation logic at some point. Too many issues caused by the driver still basically doing what it has done since cloud-beta times with some more features and waitForActions sprinkled across. If there is any more input you want to share, I'd be glad to hear it.

[apricote]

Awesome! Thanks for the quick improvement :)

I used to be a user of the project myself back in 2018-2020, its nice to see you still work on this after all these years!

[Wouter0100]

Thanks for the investigation, @apricote - and for the quick implementation of the suggested changes @JonasProgrammer!

I think we were the customer you were investigating for @apricote, and we now updated this docker-machine driver to ensure we won't hit any limits again :-).