Re: What happened with the Patch "New RAID library supporting up to six parities"
From: Ronny Egner
Date: Tue Oct 21 2014 - 09:17:36 EST
Hi Neil,
i did a short test and it works so far. Here are my results. Let me know
if you need something more:
(TL;DR: Wonderful patch. Tested with PAR6 (= six parities) and was able to
recover from losing five disks at once.)
The patches apply against 3.14.22 and btrfs-progs 3.12 but not against the
recent 3.18-rc1 and btrfs-progs > 3.12.
root@ubuntu-1204-build:~# btrfs --version
Btrfs v3.12-dirty
root@ubuntu-1204-build:~# uname -a
Linux ubuntu-1204-build 3.14.22 #3 SMP Tue Oct 21 13:00:08 CEST 2014
x86_64 x86_64 x86_64 GNU/Linux
For the tests i used a VM with 4 GB memory, two cores and 15 disks with
150 GB each. Every disk looked like this:
root@ubuntu-1204-build:~# fdisk /dev/sdi
Command (m for help): p
Disk /dev/sdi: 157.3 GB, 157286400000 bytes
81 heads, 30 sectors/track, 126419 cylinders, total 307200000 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x5b5d7269
Device Boot Start End Blocks Id System
/dev/sdi1 2048 307199999 153598976 83 Linux
File system created:
root@ubuntu-1204-build:~# mkfs.btrfs -dpar6 -L testpar6 /dev/sdh1
/dev/sdi1 /dev/sdj1 /dev/sdk1 \
/dev/sdl1 /dev/sdm1 /dev/sdn1 /dev/sdo1 /dev/sdp1
/dev/sdq1 /dev/sdr1 \
/dev/sds1 /dev/sdt1 /dev/sdu1 /dev/sdv1
Turning ON incompat feature 'extref': increased hardlink limit per file
to 65536
Turning ON incompat feature 'par3456': raid support with up to six
parities
adding device /dev/sdi1 id 2
adding device /dev/sdj1 id 3
adding device /dev/sdk1 id 4
adding device /dev/sdl1 id 5
adding device /dev/sdm1 id 6
adding device /dev/sdn1 id 7
adding device /dev/sdo1 id 8
adding device /dev/sdp1 id 9
adding device /dev/sdq1 id 10
adding device /dev/sdr1 id 11
adding device /dev/sds1 id 12
adding device /dev/sdt1 id 13
adding device /dev/sdu1 id 14
adding device /dev/sdv1 id 15
fs created label testpar6 on /dev/sdh1
nodesize 16384 leafsize 16384 sectorsize 4096 size 2.15TiB
Btrfs v3.12-dirty
Mount:
root@ubuntu-1204-build:~# mount /dev/sdh1 /mnt
Stats:
root@ubuntu-1204-build:~# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/vgroot-lvroot 26G 17G 8.4G 67% /
none 4.0K 0 4.0K 0% /sys/fs/cgroup
udev 1.6G 4.0K 1.6G 1% /dev
tmpfs 331M 1.1M 330M 1% /run
none 5.0M 0 5.0M 0% /run/lock
none 1.7G 0 1.7G 0% /run/shm
none 100M 0 100M 0% /run/user
/dev/sdh1 2.2T 2.8M 2.2T 1% /mnt
Data, single: total=8.00MiB, used=0.00
Data, PAR6: total=9.00GiB, used=995.16MiB
System, RAID1: total=8.00MiB, used=16.00KiB
System, single: total=4.00MiB, used=0.00
Metadata, RAID1: total=1.00GiB, used=65.59MiB
Metadata, single: total=8.00MiB, used=0.00
root@ubuntu-1204-build:/mnt# btrfs fi show
Label: testpar6 uuid: 79d3c5e4-74ce-4464-a509-ef666dcd9073
Total devices 15 FS bytes used 1.04GiB
devid 1 size 146.48GiB used 1.02GiB path /dev/sdh1
devid 2 size 146.48GiB used 1.00GiB path /dev/sdi1
devid 3 size 146.48GiB used 1.00GiB path /dev/sdj1
devid 4 size 146.48GiB used 1.00GiB path /dev/sdk1
devid 5 size 146.48GiB used 1.00GiB path /dev/sdl1
devid 6 size 146.48GiB used 1.00GiB path /dev/sdm1
devid 7 size 146.48GiB used 1.00GiB path /dev/sdn1
devid 8 size 146.48GiB used 1.00GiB path /dev/sdo1
devid 9 size 146.48GiB used 1.00GiB path /dev/sdp1
devid 10 size 146.48GiB used 1.00GiB path /dev/sdq1
devid 11 size 146.48GiB used 1.00GiB path /dev/sdr1
devid 12 size 146.48GiB used 2.00GiB path /dev/sds1
devid 13 size 146.48GiB used 2.00GiB path /dev/sdt1
devid 14 size 146.48GiB used 1.01GiB path /dev/sdu1
devid 15 size 146.48GiB used 1.01GiB path /dev/sdv1
Metadata and data still âsingleâ? Bug? Nevermind - lets convert it:
root@ubuntu-1204-build:/mnt# btrfs balance start -mconvert=raid1 /mnt
Done, had to relocate 4 out of 6 chunks
root@ubuntu-1204-build:/mnt# btrfs fi df /mnt
Data, single: total=8.00MiB, used=0.00
Data, PAR6: total=9.00GiB, used=1.02GiB
System, RAID1: total=32.00MiB, used=16.00KiB
Metadata, RAID1: total=1.00GiB, used=67.83MiB
root@ubuntu-1204-build:/mnt# btrfs balance start -dconvert=par6 /mnt
Done, had to relocate 2 out of 4 chunks
root@ubuntu-1204-build:/mnt# btrfs fi df /mnt
Data, PAR6: total=9.00GiB, used=1.02GiB
System, RAID1: total=32.00MiB, used=16.00KiB
Metadata, RAID1: total=1.00GiB, used=68.72MiB
OK now lets see what happens if we remove on device. Save a MD5SUM before:
root@ubuntu-1204-build:/mnt# md5sum linux-3.14.22.tar
80af37cdfb2fa2239f79597c914a8c73 linux-3.14.22.tar
(Removed one disk and replace it with a brand new, empty one)
root@ubuntu-1204-build:~# mount /dev/sdh1 /mnt
mount: wrong fs type, bad option, bad superblock on /dev/sdh1,
missing codepage or helper program, or other error
In some cases useful info is found in syslog - try
dmesg | tail or so
root@ubuntu-1204-build:~# mount /dev/sdh1 /mnt -o degraded
root@ubuntu-1204-build:~#
root@ubuntu-1204-build:~# btrfs fi show
Label: testpar6 uuid: 79d3c5e4-74ce-4464-a509-ef666dcd9073
Total devices 15 FS bytes used 31.42GiB
devid 1 size 146.48GiB used 4.00GiB path /dev/sdh1
devid 2 size 146.48GiB used 5.00GiB path /dev/sdi1
devid 3 size 146.48GiB used 4.00GiB path /dev/sdj1
devid 4 size 146.48GiB used 4.00GiB path /dev/sdk1
devid 5 size 146.48GiB used 4.00GiB path /dev/sdl1
devid 6 size 146.48GiB used 5.00GiB path /dev/sdm1
devid 7 size 146.48GiB used 4.03GiB path /dev/sdn1
devid 8 size 146.48GiB used 4.00GiB path /dev/sdo1
devid 9 size 146.48GiB used 4.00GiB path /dev/sdp1
devid 10 size 146.48GiB used 4.00GiB path /dev/sdq1
devid 11 size 146.48GiB used 4.00GiB path /dev/sdr1
devid 12 size 146.48GiB used 4.03GiB path /dev/sds1
devid 13 size 146.48GiB used 4.00GiB path /dev/sdt1
devid 14 size 146.48GiB used 4.00GiB path /dev/sdu1
devid 15 size 146.48GiB used 4.00GiB path
Lets replace the faulty disk:
root@ubuntu-1204-build:~# btrfs device add /dev/sdv1 /mnt
root@ubuntu-1204-build:~# btrfs device delete missing /mnt
In /var/log/syslog:
[ 191.442050] BTRFS warning (device sdk1): devid 15 missing
[ 581.367659] sdv: sdv1
[ 598.009968] BTRFS: device label testpar6 devid 16 transid 63 /dev/sdv1
[ 614.679654] BTRFS info (device sdk1): relocating block group
40865103872 flags 4097
[ 657.889822] BTRFS info (device sdk1): found 64 extents
[ 659.190497] BTRFS info (device sdk1): found 64 extents
[ 659.247765] BTRFS info (device sdk1): relocating block group
31201427456 flags 4097
[ 861.359599] BTRFS info (device sdk1): found 132 extents
[ 862.875521] BTRFS info (device sdk1): found 132 extents
[ 862.973499] BTRFS info (device sdk1): relocating block group
11874074624 flags 4097
After the âdelete missingâ
Label: testpar6 uuid: 79d3c5e4-74ce-4464-a509-ef666dcd9073
Total devices 15 FS bytes used 31.42GiB
devid 1 size 146.48GiB used 4.00GiB path /dev/sdh1
. . .
devid 14 size 146.48GiB used 4.00GiB path /dev/sdu1
devid 16 size 146.48GiB used 4.00GiB path /dev/sdv1
The md5 checksum is still correct:
root@ubuntu-1204-build:/mnt# md5sum linux-3.14.22.tar
80af37cdfb2fa2239f79597c914a8c73 linux-3.14.22.tar
Hardcore test: PAR6 = 6 parities. LetÂs see what happens if i remove five
disks and replace with with empty ones.
Before i did that the metadata format was converted to PAR6 as well:
root@ubuntu-1204-build:~# btrfs fi df /mnt/
Data, PAR6: total=36.00GiB, used=31.32GiB
System, PAR6: total=144.00MiB, used=16.00KiB
Metadata, PAR6: total=1.12GiB, used=101.81MiB
root@ubuntu-1204-build:~# mount /dev/sdn1 /mnt/ -o degraded
root@ubuntu-1204-build:~# btrfs fi show
Label: testpar6 uuid: 79d3c5e4-74ce-4464-a509-ef666dcd9073
Total devices 15 FS bytes used 31.42GiB
devid 1 size 146.48GiB used 4.00GiB path
devid 2 size 146.48GiB used 5.00GiB path
devid 3 size 146.48GiB used 4.00GiB path
devid 4 size 146.48GiB used 4.00GiB path
devid 5 size 146.48GiB used 4.00GiB path
devid 6 size 146.48GiB used 5.00GiB path /dev/sdm1
devid 7 size 146.48GiB used 4.03GiB path /dev/sdn1
devid 8 size 146.48GiB used 4.00GiB path /dev/sdo1
devid 9 size 146.48GiB used 4.00GiB path /dev/sdp1
devid 10 size 146.48GiB used 4.00GiB path /dev/sdq1
devid 11 size 146.48GiB used 4.00GiB path /dev/sdr1
devid 12 size 146.48GiB used 4.03GiB path /dev/sds1
devid 13 size 146.48GiB used 4.00GiB path /dev/sdt1
devid 14 size 146.48GiB used 4.00GiB path /dev/sdu1
devid 16 size 146.48GiB used 4.00GiB path /dev/sdv1
Now letÂs bring it back in shape and add five new, empty disks:
btrfs device add /dev/sdh1 /dev/sdi1 /dev/sdj1 /dev/sdk1 /dev/sdl1 /mnt
btrfs delete missing
<<wait>>
root@ubuntu-1204-build:~# btrfs fi show
Label: testpar6 uuid: 79d3c5e4-74ce-4464-a509-ef666dcd9073
Total devices 15 FS bytes used 1.09GiB
devid 6 size 146.48GiB used 2.14GiB path /dev/sdm1
devid 7 size 146.48GiB used 2.14GiB path /dev/sdn1
devid 8 size 146.48GiB used 2.14GiB path /dev/sdo1
devid 9 size 146.48GiB used 2.14GiB path /dev/sdp1
devid 10 size 146.48GiB used 2.14GiB path /dev/sdq1
devid 11 size 146.48GiB used 2.14GiB path /dev/sdr1
devid 12 size 146.48GiB used 2.14GiB path /dev/sds1
devid 13 size 146.48GiB used 2.14GiB path /dev/sdt1
devid 14 size 146.48GiB used 2.14GiB path /dev/sdu1
devid 16 size 146.48GiB used 2.14GiB path /dev/sdv1
devid 17 size 146.48GiB used 2.14GiB path /dev/sdh1
devid 18 size 146.48GiB used 2.14GiB path /dev/sdi1
devid 19 size 146.48GiB used 2.14GiB path /dev/sdj1
devid 20 size 146.48GiB used 2.14GiB path /dev/sdk1
devid 21 size 146.48GiB used 2.14GiB path /dev/sdl1
And now the checksum:
root@ubuntu-1204-build:/mnt# md5sum linux-3.14.22.tar
80af37cdfb2fa2239f79597c914a8c73 linux-3.14.22.tar
Checksum matches!
So.. this looks *very* good to me.
Mit freundlichen GrÃÃen
Ronny Egner
--
Ronny Egner
Oracle Certified Master 11g (OCM)
Mobile: +49 170 8139903
EMail: ronnyegner@xxxxxxxxxxxxxxxxxxxxxxxx
<mailto:roonnyegner@xxxxxxxxxxxxxxxxxxxxxxxx>
Am 21.10.14 09:27 schrieb "NeilBrown" unter <neilb@xxxxxxx>:
>On Tue, 21 Oct 2014 06:33:47 +0000 Ronny Egner
><ronnyegner@xxxxxxxxxxxxxxxxxxxxxxxx> wrote:
>
>> Dear All,
>>
>> i was wondering what happened with the patch posted by Andrea Mazzoleni
>> back in Februrary 2014 (this Thread:
>> http://thread.gmane.org/gmane.linux.kernel/1654735).
>>
>> Why washÂt it added to the code? Something missing/wrong?
>>
>> In my opinion the posted patch is awesome and would enable a unique
>> feature that no other UNIX-like operating system currently has.
>>
>
>Could you report your test results please.
>
>NeilBrown