[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-bcache
Subject:    Re: Large latency with bcache for Ceph OSD
From:       "Norman.Kern" <norman.kern () gmx ! com>
Date:       2021-03-02 5:30:21
Message-ID: e67ea64d-7e36-761d-0e28-cc0d1012c3aa () gmx ! com
[Download RAW message or body]


On 2021/2/26 =E4=B8=8B=E5=8D=885:54, Coly Li wrote:
> On 2/26/21 4:57 PM, Norman.Kern wrote:
> [snipped]
>>> You may try to trigger a gc by writing to
>>> sys/fs/bcache/<cache-set-uuid>/internal/trigger_gc
>>>
>> When all cache had written back, I triggered gc, it recalled.
>>
>> root@WXS0106:~# cat /sys/block/bcache0/bcache/cache/cache_available_per=
cent
>> 30
>>
>> root@WXS0106:~# echo 1 > /sys/block/bcache0/bcache/cache/internal/trigg=
er_gc
>> root@WXS0106:~# cat /sys/block/bcache0/bcache/cache/cache_available_per=
cent
>> 97
>>
>> Why must I trigger gc manually? Is not a default action of bcache-gc th=
read? And I found it can only work when all dirty data written back.
>>
> 1, GC is automatically triggered after some mount of data consumed. I
> guess it is just not about time in your situation.
>
> 2, Because the gc will shrink all cached clean data, which is very
> unfriendly for read-intend workload. Therefore gc_after_writeback is
> defaulted as 0, when this sysfs file content set to 1, a gc will trigger
> after the writeback accomplished.

I made a test again and get more infomation:

root@WXS0089:~# cat /sys/block/bcache0/bcache/dirty_data
0.0k
root@WXS0089:~# lsblk /dev/sda
NAME=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 MAJ:MIN RM=C2=A0=C2=A0 SIZE RO TYPE MOU=
NTPOINT
sda=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 8:0=C2=A0=C2=A0=C2=A0 =
0 447.1G=C2=A0 0 disk
`-bcache0 252:0=C2=A0=C2=A0=C2=A0 0=C2=A0 10.9T=C2=A0 0 disk
root@WXS0089:~# cat /sys/block/sda/bcache/priority_stats
Unused:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 1%
Clean:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 29%
Dirty:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 70%
Metadata:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 0%
Average:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 49
Sectors per Q:=C2=A0 29184768
Quantiles:=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 [1 2 3 5 6 8 9 11 13 14 16 19 21 =
23 26 29 32 36 39 43 48 53 59 65 73 83 94 109 129 156 203]
root@WXS0089:~# cat /sys/fs/bcache/066319e1-8680-4b5b-adb8-49596319154b/in=
ternal/gc_after_writeback
1
You have new mail in /var/mail/root
root@WXS0089:~# cat /sys/fs/bcache/066319e1-8680-4b5b-adb8-49596319154b/ca=
che_available_percent
28

I read the source codes and found if cache_available_percent > 50, it shou=
ld wakeup gc while doing writeback, but it seemed not work right.

>
> Coly Li
>
>
>
>
[prev in list] [next in list] [prev in thread] [next in thread]