collector/diskstats: add device-mapper backing device info metric

[Maharacha]

We have this use case where we need to see what device is backing a device mapper.

Example:

node_disk_device_mapper_backing_device_info{instance="foo-bar",device="dm-0",backing_device="nvme0n1"} 1
node_disk_device_mapper_backing_device_info{instance="foo-bar",device="dm-1",backing_device="nvme0n1"} 1
node_disk_device_mapper_backing_device_info{instance="foo-bar",device="dm-2",backing_device="dm-0"} 1
node_disk_device_mapper_backing_device_info{instance="foo-bar",device="dm-2",backing_device="dm-1"} 1
node_disk_device_mapper_backing_device_info{instance="foo-bar",device="dm-3",backing_device="dm-2"} 1
node_disk_device_mapper_backing_device_info{instance="foo-bar",device="dm-4",backing_device="dm-2"} 1
node_disk_device_mapper_backing_device_info{instance="foo-bar",device="dm-5",backing_device="dm-2"} 1

Together with the already existing metric node_disk_device_mapper_info, it's now possible to map the backing devices to volume groups.

[erik78se]

This is a great addition! Well needed! PV:s are critical to surface like this. +1 from me.

[erik78se]

@SuperQ - is there anything we need to do to get this into the exporter main ?

[nicolastakashi]

@erik78se could you confirm the e2e check is green by running: go build && ./end-to-end-test.sh -u

[nicolastakashi]

nit: IIRC other places we are following this style of validation.

Suggested change

	if !os.IsNotExist(err) {
	if errors.Is(err, os.ErrNotExist)

[Maharacha]

Updated!

[Maharacha]

@erik78se could you confirm the e2e check is green by running: go build && ./end-to-end-test.sh -u

@nicolastakashi I have updated the fixture so that ./end-to-end-test.sh is green now. 785de74

[nicolastakashi]

@Maharacha can you fix CI?

[Maharacha]

@Maharacha can you fix CI?

I have added this d7a6d33

Can you approve workflows?

[ArthurSens]

Interesting case here, if I understand it correctly, dm-1/2/3/4/5 are all created from the backing_device dm-0, and dm-0 is backed by nvme0n1p3. So operations in the disk dm-1/2/3/4/5 are all handled by nvme0n1p3 in reality, right?

How would a query to find the real backing device look like in a scenario like this that requires moving the layer twice, or maybe even more in other scenarios?