lsblk.py: add devid in output

[Ythogtha]

Add a function to get disk/by-id names from a device name, and add it to the output

Any resemblance with a previous PR is not completely random.

Performance tests are on par with the previous PR, which is : not good enough.
New version will call once the udev database, and get all data readily available, which is way better.

Here is a comparison of the four methods, times are for querying sequentially the three hosts of the pool.
Pool (l1, l2, l3) is the corresponding pool on XOlab.
Pool (yp1, yp2, yp3) is a virtual home pool.
Using command: xe host-call-plugin host-uuid=<uuid> plugin=lsblk.py fn=list_block_devices.

|------+-------+-------+---------+----------|
| host | org   | find  | udevadm | exportdb |
|------+-------+-------+---------+----------|
| l1   | 1,220 | 4,024 | 4,452   | 1,747    |
| l2   | 1,449 | 3,806 | 4,248   | 1,566    |
| l3   | 1,352 | 4,016 | 4,465   | 1,715    |
|------+-------+-------+---------+----------|
| yp1  | 0,198 | 1,140 | 1,108   | 0,274    |
| yp2  | 0,248 | 1,082 | 1,068   | 0,343    |
| yp3  | 0,243 | 1,127 | 1,090   | 0,322    |
|------+-------+-------+---------+----------|

There are other differences between the three methods

• find returns [""] of no link are found.
• udevadm (which makes a udevadm call for each device), filter results for /dev/disk/by-id/
• exportdb (which calls udevadm once at the beginning), returns all links in /dev/disk/by-*/

The scope of the links returned could be tuned to limit to by-id, or not, performances wouldn't be altered.

[psafont]

Hi, I'm looking at the PR. I've refreshed the CI to test the changes on Python 3.6 on master, and so I've rebased your commits on it

[Ythogtha]

Hi, I'm looking at the PR. I've refreshed the CI to test the changes on Python 3.6 on master, and so I've rebased your commits on it

Ok!
And please, do not hesitate to give your insight, if any.

[psafont]

Looks OK from what I've seen, can you squash the commits?

[Wescoeur]

Ping for review @Nambrok

[Nambrok]

I have not identified problems with the code, by I do think we only want the by-id one to identify disks.
In this configuration, we also have FS and partition path, those could change.

[Ythogtha]

I have not identified problems with the code, by I do think we only want the by-id one to identify disks. In this configuration, we also have FS and partition path, those could change.

Done, with a simple way to change behavior in the future, and include other by-something, like by-label for example.