Make it easier to plug in other providers and add a mock provider

[jamielinux]

mock provider

Before I submitted the Add post_plan_delay option PR, I wanted a way to test it locally. So I created a "mock" provider.

Otherwise I wasn't really sure how to easily test it without using actual hetzner credentials, unless I missed something!

$ ./flipper --config mock-flipper.example.yaml monitor
time=2025-12-08T18:26:22.268Z level=DEBUG msg="Starting resource watcher." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock
time=2025-12-08T18:26:22.268Z level=INFO msg="Resources changed, updating healthkeeper state." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock
time=2025-12-08T18:26:22.269Z level=DEBUG msg="Server health check completed." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock server_id=2 server_name=mock-server-2 check_id=health__ipv6 server_state=unknown rise=0 fall=1 duration=338.434µs
time=2025-12-08T18:26:22.269Z level=ERROR msg="Actions required." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock plan="Plan{ 100 -> 1 }" plan_id=c8418003-be35-4f4e-acf0-1c2d5049b8e6 unhealthy_floating_ips=[]
time=2025-12-08T18:26:22.269Z level=DEBUG msg="Server health check completed." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock server_id=1 server_name=mock-server-1 check_id=health__ipv6 server_state=unknown rise=0 fall=1 duration=266.671µs
time=2025-12-08T18:26:22.269Z level=ERROR msg="Actions required." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock plan="Plan{ 100 -> 1 }" plan_id=ebcc4c60-aeb0-4a13-81fd-2a6390e40d5d unhealthy_floating_ips=[]
time=2025-12-08T18:26:22.269Z level=INFO msg="Executing plan." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock plan_id=c8418003-be35-4f4e-acf0-1c2d5049b8e6
time=2025-12-08T18:26:22.269Z level=INFO msg="Floating IP assigned." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock plan_id=c8418003-be35-4f4e-acf0-1c2d5049b8e6 floating_ip_id=100 server_id=1
time=2025-12-08T18:26:22.269Z level=INFO msg="Plan executed successfully." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock plan_id=c8418003-be35-4f4e-acf0-1c2d5049b8e6 num_unhealthy_servers=0
time=2025-12-08T18:26:27.269Z level=INFO msg="Ignoring stale action with sequence lower than min sequence." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock sequence=1 min_sequence=3
time=2025-12-08T18:26:27.269Z level=INFO msg="Resources changed, updating healthkeeper state." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock
time=2025-12-08T18:26:27.269Z level=DEBUG msg="Server health check completed." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock server_id=2 server_name=mock-server-2 check_id=health__ipv4 server_state=unknown rise=0 fall=1 duration=1.000436195s
time=2025-12-08T18:26:27.269Z level=DEBUG msg="No actions required." version=<unknown> service=flipper group="Local Test" group_id=test_group provider=mock

HetznerID

I noticed that HetznerID could be more generic, so I changed that to ServerID (or something else appropriate).

Wait() called on empty errgroup

I think there's a typo here:

flipper/monitor/monitor.go

Lines 68 to 80 in 8ec4926

	errgrp := errgroup.Group{}
	errggrp, ctx := errgroup.WithContext(ctx)
	for _, group := range w.groups {
	group := group
	errggrp.Go(func() error {
	return group.Start(ctx)
	})
	}
	//nolint:wrapcheck // we wrap the error in the subroutines.
	return errgrp.Wait()
	}

This causes a wait on the empty errgrp (instead of errggrp).

I noticed this because running ./flipper --config mock-flipper.example.yaml monitor was exiting immediately with no logs and 0 exit status.

I think this doesn't really affect production use:

• Watch() returns immediately and the caller thinks monitoring is completed.
• Monitoring keeps running because the Group.Start() goroutines are still blocking in their for-select loops (and their child goroutines like g.healthkeeper.Start() are also still running).
• server.ListenAndServe() keeps the process alive, so those orphaned goroutines continue working sort of by accident.

It's a small fix, but I also added a test.

[gzuidhof]

The reason for the HetznerID is that in Hetzner these IDs are int64s, but at a different provider it may be something else (e.g. I wouldn't be surprised if it's a string in most places).

I would propose to add a field per provider, so if we would add support for OVH (or whichever), the ID() function would have a switch on the ProviderName to decide which one to return. (That's also why ID() returns a string, which any ID should be castable to).

[jamielinux]

Ahhh right of course, thanks! 🚧

[jamielinux]

I think I've implemented your suggestion 🤔 Assuming I interpreted your approach correctly!

[gzuidhof]

Looks great :)

Could you update the PR name before merging (just for bookkeeping so it's not confusing in the changelog)