discovery.Manager.runUpdater() only needs to run once

[rainpavilion]

A provider runs runUpdater()，but they do same things, it maybe a bug.

[simonpasquier]

Ah, you might have spotted "interesting"... I agree that a new runUpdater() goroutine is spawned for every SD provider while only one instance is needed AFAICT. This could explain SD issues like #4124.

prometheus/discovery/manager.go

Lines 127 to 136 in 46fb407

	func (m *Manager) startProvider(ctx context.Context, poolKey poolKey, worker Discoverer) {
	ctx, cancel := context.WithCancel(ctx)
	updates := make(chan []*targetgroup.Group)
	m.discoverCancel = append(m.discoverCancel, cancel)
	go worker.Run(ctx, updates)
	go m.runProvider(ctx, poolKey, updates)
	go m.runUpdater(ctx)
	}

@krasi-georgiev?

[krasi-georgiev]

yes we only need one runUpdater so the logic can be improved there.
We added this just to throttle the updates to no more than one per 5 * time.Second.
The only side effect of the current code is that it will run the updates more often than desired, but shouldn't cause missing updates as in #4124.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.