Process BTRFS and whole-disk filesystem stats

[arkadijs]

XFS users might complain, but we - Docker users - use BTRFS. :)
Typical usage of CoreOS (on AWS) is to format whole device to BTRFS and mount it on /var/lib/docker, thus whole-disk stats must also be collected.

[vishh]

"TestGetDiskStatsMap" unit test is failing. Can you address the issue?

[arkadijs]

Sorry for that, fixed.
I can force-push as a single commit if necessary.

[vmarmol]

LGTM, yes please. One commit would be nice :)

[arkadijs]

Also added device-mapper to regex. Works for BTRFS volumes on LVM as they appears as /dev/dm-* in /proc/mounts even if mounted as /dev/vg/lv.
Not a complete solution because ext4 on LVM appears as intended, ie. /dev/vg/lv and thus does not match /proc/diskstats. No idea why such difference.

[arkadijs]

I checked it on Docker host running with devicemapper storage driver. It doesn't help to collect the stats, but not making this worse either:

{
  "device": "/dev/mapper/docker-202:1-922914-e6128e016bdac12e73c36d70659e4984159878c3c33e1e3b962891185aaeb525",
  "capacity": 10434662400,
  "usage": 651108352,
  "reads_completed": 0,
  "reads_merged": 0,
  "sectors_read": 0,
  "read_time": 0,
  "writes_completed": 0,
  "writes_merged": 0,
  "sectors_written": 0,
  "write_time": 0,
  "io_in_progress": 0,
  "io_time": 0,
  "weighted_io_time": 0
},

[vishh]

LGTM. @vmarmol any other comments?

[cadvisorJenkinsBot]

Can one of the admins verify this patch?

[vmarmol]

LGTM