3.17: Yay! New sizes in df -h

Hi!

Now I get this with 3.17-rc2:

merkaba:~> LANG=C df -hT / /home

Filesystem     Type   Size  Used Avail Use% Mounted on
/dev/dm-5      btrfs   30G   17G   23G  43% /
/dev/dm-0      btrfs  160G  129G   59G  69% /home

/: 17+23 = 40 GiB, compressed 10 GiB away with lzo? /home: 59+129 = 188 GiB, compressed 28 GiB away with lzo?

This is BTRFS RAID 1 on Dual SSD:

merkaba:~> btrfs fi df /
Data, RAID1: total=25.00GiB, used=15.33GiB System, RAID1: total=8.00MiB, used=16.00KiB Metadata, RAID1: total=2.00GiB, used=677.33MiB unknown, single: total=240.00MiB, used=0.00

merkaba:~> btrfs fi df /home
Data, RAID1: total=154.97GiB, used=125.70GiB System, RAID1: total=32.00MiB, used=48.00KiB Metadata, RAID1: total=5.00GiB, used=2.71GiB unknown, single: total=512.00MiB, used=0.00

Ciao,

Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7

Hmm, I'm not seeing such a large discrepancy between Used+Avail and Size.

No compression, all volumes populated with send/receive and contain same data.

3.16.1

# df -h
Filesystem               Size  Used Avail Use% Mounted on
/dev/sdb                  40G  2.8G   37G   7% /home        # single 40GB device
/dev/sdc                  40G  5.5G   33G  15% /mnt	    # two 20GB devices, raid1

Reboot 3.17.0rc2

# df -h
Filesystem               Size  Used Avail Use% Mounted on
/dev/sdb                  40G  2.8G   37G   7% /home        # single 40GB device
/dev/sdc                  20G  2.8G   17G  15% /mnt	    # two 20GB devices, raid1

======
sde is two 20GB devices, raid1, -o compress, populated the same as above w/ send/receive

3.16.1
Filesystem Size Used Avail Use% Mounted on
/dev/sde 40G 3.8G 35G 10% /mnt

3.17.0rc2
Filesystem Size Used Avail Use% Mounted on
/dev/sde 20G 2.0G 18G 10% /mnt

In the last example, Used went from 2.8G to 2.0G due to compression. But with or without compression, Used+Avail=Size.

Chris Murphy

Am Mittwoch, 27. August 2014, 10:54:56 schrieb Chris Murphy:

I see.

Well the /home has a large maildir and lots of git checkouts with highly compressible text files.

My data directory for larger files which is only on the mSATA SSD doesn´t have this discrepancy:

merkaba:~> LANG=C df -hT /daten

Filesystem              Type   Size  Used Avail Use% Mounted on
/dev/mapper/msata-daten btrfs  200G  183G   17G  92% /daten

merkaba:~> btrfs fi df /daten
Data, single: total=187.01GiB, used=182.05GiB
System, single: total=4.00MiB, used=48.00KiB
Metadata, single: total=1.01GiB, used=288.09MiB
unknown, single: total=112.00MiB, used=0.00

I also use compress=lzo on it, although it I wonder whether its necessary as I think the Crucial m500 mSATA SSD firmware compresses itself while the Intel SSD 320 SATA SSD firmware does not. But still even with compression in the firmware, it BTRFS compresses I can use more of the capacity. I wonder whether this can harm the reliability of the SSD, if its firmware relies on at least partly compressible data.

Well I wouldn´t expect to see discrepancies without compressed data. But on the other hand I did not dig into how the new calculation actually works so far.

Hmmm, I see. Then it might not be compression at all what explains the differences I see?

Any other explainations?

Ciao,

Martin 'Helios' Steigerwald - http://www.Lichtvoll.de
GPG: 03B0 0D6C 0040 0710 4AFA  B82F 991B EAAC A599 84C7

Seems unlikely because application binaries aren't very compressible, and user videos/images aren't compressible at all because they're already highly compressed.

I don't know. At the very least I'd repost with bug in caps in your subject, and point out the Used+Avail≠Size discrepancy, and include strace df results. I think your results are sufficiently confusing to consider it a bug.

Chris Murphy