Add devicemapper GC for persistent rootfs

[ivan4th]

Upon GC at the Virtlet startup, the virtual block devices that are
created by Virtlet are recognized by name and the sector 0 of underlying
block device (Virtlet magic header). If they're not related to any
active domain, they're removed (note that this doesn't mean the data on
the device is affected in any way).

---

This change is 

[jellonek]

Reviewed 10 of 10 files at r1.
Reviewable status: 0 of 2 approvals obtained (waiting on @ivan4th)

---

pkg/libvirttools/gc.go, line 275 at r1 (raw file):

		idsInUse[id] = true
	}
	ldh := blockdev.NewLogicalDeviceHandler(v.Commander(), devPath, sysfsPath)

Maybe this could be done during creation of VirtualizationTool once, stored in its structure, then could be used there and also could be returned by devHandler?

[lukaszo]

Reviewable status: 1 change requests, 0 of 2 approvals obtained (waiting on @ivan4th)

---

pkg/libvirttools/gc.go, line 284 at r1 (raw file):

	for _, dmName := range dmNames {
		if !strings.HasPrefix(dmName, blockdev.VirtletLogicalDevicePrefix) {
			panic("bad dmname " + dmName)

why panic?

[jellonek]

Reviewable status: 2 change requests, 0 of 2 approvals obtained (waiting on @ivan4th)

---

pkg/libvirttools/gc.go, line 284 at r1 (raw file):

Previously, lukaszo (Łukasz Oleś) wrote…

why panic?

I'm reading that ListVirtletLogicalDevices returned a device which has virtlet magic but wrong name. This should never happened, unless there was breaking api virtlet pod upgrade which would include change of virtlet logical device prefix.

[lukaszo]

Reviewable status: 2 change requests, 0 of 2 approvals obtained (waiting on @ivan4th)

---

pkg/libvirttools/gc.go, line 284 at r1 (raw file):

Previously, jellonek (Piotr Skamruk) wrote…

I'm reading that ListVirtletLogicalDevices returned a device which has virtlet magic but wrong name. This should never happened, unless there was breaking api virtlet pod upgrade which would include change of virtlet logical device prefix.

It's gc. I think that device can just be skipped with a big warning message.

[ivan4th]

Reviewable status: 2 change requests, 0 of 2 approvals obtained (waiting on @ivan4th and @jellonek)

---

pkg/libvirttools/gc.go, line 275 at r1 (raw file):

Previously, jellonek (Piotr Skamruk) wrote…

Maybe this could be done during creation of VirtualizationTool once, stored in its structure, then could be used there and also could be returned by devHandler?

I thought about this. What stopped me from this approach is that with gc.go being a set of VirtualizationTool methods, VirtualizationTool itself becomes an instance of "God Object" antipattern. Because of this, I'd prefer not to add extra coupling between GC methods and other parts of VirtualizationTool.

---

pkg/libvirttools/gc.go, line 284 at r1 (raw file):

Previously, lukaszo (Łukasz Oleś) wrote…

It's gc. I think that device can just be skipped with a big warning message.

panic is used here because ListVirtletLogicalDevices should only return Virtlet-related devices that MUST include the prefix in their names. If that's not the case, this means some severe bug in Virtlet code.

[jellonek]

Reviewable status: 2 change requests, 0 of 2 approvals obtained (waiting on @jellonek and @lukaszo)

---

pkg/libvirttools/gc.go, line 275 at r1 (raw file):

Previously, ivan4th (Ivan Shvedunov) wrote…

I thought about this. What stopped me from this approach is that with gc.go being a set of VirtualizationTool methods, VirtualizationTool itself becomes an instance of "God Object" antipattern. Because of this, I'd prefer not to add extra coupling between GC methods and other parts of VirtualizationTool.

That's solid reason with which I fully agree.

[pigmej]

Reviewed 10 of 10 files at r1.
Reviewable status: 1 change requests, 1 of 2 approvals obtained (waiting on @ivan4th and @lukaszo)