Hello list!
During 3.9 I had dead-lock problems with autodefrag. The system either froze
or showed some kernel oopses randomly while working with the system. So I
turned off autodefrag and the problem was cured.
But that radically slowed my system down over time. My desktop programs now
take almost 10 times as long to start in comparision to 3.8 and beginning of
3.9 before I hit the bug. This also affects opening file dialogs, browsing
directories with the graphical filebrowser, hell: Even right-click menus
sometimes take ages to open while the harddisk LED happily brightens my
room.
I have a three device btrfs filesystem (rotational, 3x 1TB) with non-forced
lzo compression. Before, startup of KDE programs was almost immediatly after
clicking on the icons - even after login.
Now, since 3.10 the freeze bug seems to be fixed and I turned autodefrag
back on. But the performance simply does not return (I did not expect the
effect to be immediatly). Then, I've turned on skinny extent which made a
little subjective difference. But still the system is far from the
performance before I turned off autodefrag.
I'm not sure if there is still some performance regression bug in 3.10 or if
I have to run some defragmentation. I've run a defrag on each directory with
find / -xdev -type d -print0 | xargs -0 btrfs fi defrag -v
and it helped (apparently only) a little but after a few days of usage the
system returns to its newly developed slowness. :-(
I've changed nr_requests in the I/O scheduler to 4096 which also gives a
subjective small speed boost during I/O intensive loads. The system has 16
GB of memory with at least 8 GB cached most of the time. So it is no memory
pressure problem.
Unfortunatly, the defrag process does not finish. Somewhere in the process
the system hard freezes with no chance to get a backtrace from kernel. At
least it works for almost the complete filesystem so it should be
sufficient.
A friend of mine with a single SSD reports similar slowdown issues lately
and it makes him almost wanting to return to Ext4.
My btrfs system was so blazing fast in the beginning, and now when it starts
to get older (almost 2 years now), it comes to a crawl. And it started with
turning off autodefrag and never returned to its old speed when I turned it
back on.
I'm not doing any snapshots on this system which would explain increased
fragmentation.
Are there any known issues? Are there suggestions how to investigate or fix
it?
Mount options:
/dev/sda3 on / type btrfs (rw,relatime,compress=lzo,space_cache,autodefrag)
Label: 'system' uuid: d2bb232a-2e8f-4951-8bcc-97e237f1b536
Total devices 3 FS bytes used 1.07TB
devid 3 size 927.26GB used 428.03GB path /dev/sdc3
devid 2 size 927.26GB used 428.03GB path /dev/sdb3
devid 1 size 927.26GB used 429.07GB path /dev/sda3
Data, RAID0: total=1.15TB, used=1.06TB
System, RAID1: total=64.00MB, used=96.00KB
System: total=4.00MB, used=0.00
Metadata, RAID1: total=56.00GB, used=4.46GB
Thanks,
Kai
--
To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
