uevent: should respawn the uevent monitor

[Vicente-Cheng]

Problem:
The udev monitor will not work with any errors.

Solution:
We monitor the error and try to respawn it.

Related Issue:
harvester/harvester#4925

Test plan:
Because the original error is hard to reproduce, I added another inject error here.

1. set the parameter inject-udev-monitor-error as true.
2. try to add a disk and provision it should work

[tserong]

I found two small typos, other than that it looks fine (i.e. logic seems sensible).

I was going to say that the thing I don't understand is why it terminates in the first place if there's an error (our code has an infinite loop, right?) but then I went looking at

node-disk-manager/vendor/github.com/pilebones/go-udev/netlink/conn.go

Lines 168 to 175 in 44f601c

	default:
	_, buf, err := c.msgPeek()
	if err != nil {
	errs <- fmt.Errorf("Unable to check available uevent, err: %w", err)
	break loop // stop iteration in case of error
	}
	bufToRead <- buf
	}

and discovered the goroutine in Monitor will just stop if there's any kind of error.

[Vicente-Cheng]

Hi @tserong, thanks for the review!

I was going to say that the thing I don't understand is why it terminates in the first place if there's an error (our code has an infinite loop, right?) but then I went looking at ...

As you read the go-udev behavior, they will break loop with any error.
That means even if we do not stop the goroutine, we can not receive any udev event.

So that is why we need to respawn it (means creating a new monitor for incoming udev event)

[Vicente-Cheng]

@Mergifyio backport v0.6.x

[mergify]

backport v0.6.x

✅ Backports have been created

• #77 uevent: should respawn the uevent monitor (backport #76) has been created for branch v0.6.x

[Vicente-Cheng]

@Mergifyio backport v0.5.x

[mergify]

backport v0.5.x

✅ Backports have been created

• #78 uevent: should respawn the uevent monitor (backport #76) has been created for branch v0.5.x