All of lore.kernel.org
 help / color / mirror / Atom feed
* MD Array 'stat' File - Sectors Read
@ 2020-03-30 20:54 Marc Smith
  2020-04-09  7:11 ` Song Liu
  0 siblings, 1 reply; 4+ messages in thread
From: Marc Smith @ 2020-03-30 20:54 UTC (permalink / raw)
  To: linux-raid

Hi,

Apologies in advance, as I'm sure this question has been asked many
times and there is a standard answer, but I can't seem to find it on
forums or this mailing list.

I've always observed this behavior using 'iostat', when looking at
READ throughput numbers, the value is about 4 times more than the real
throughput number. Knowing this, I typically look at the member
devices to determine what throughput is actually being achieved (or
from the application driving the I/O).

Looking at the sectors read field in the 'stat' file for an MD array
block device:
# cat /sys/block/md127/stat && sleep 1 && cat /sys/block/md127/stat
93591416        0 55082801792        0       93        0        0
  0        0        0        0        0        0        0        0
93608938        0 55092996456        0       93        0        0
  0        0        0        0        0        0        0        0

55092996456 - 55082801792 = 10194664
10194664 * 512 = 5219667968
5219667968 / 1024 / 1024 = 4977

This device definitely isn't doing 4,977 MiB/s. So now my curiosity is
getting to me: Is this just known/expected behavior for the MD array
block devices? The numbers for WRITE sectors is always accurate as far
as I can tell. Or something configured strangely on my systems?

I'm using vanilla Linux 5.4.12.


Thanks,

Marc

^ permalink raw reply	[flat|nested] 4+ messages in thread
* Re: MD Array 'stat' File - Sectors Read
  2020-03-30 20:54 MD Array 'stat' File - Sectors Read Marc Smith
@ 2020-04-09  7:11 ` Song Liu
  2020-04-11  4:59   ` Marc Smith
  0 siblings, 1 reply; 4+ messages in thread
From: Song Liu @ 2020-04-09  7:11 UTC (permalink / raw)
  To: Marc Smith; +Cc: linux-raid

On Mon, Mar 30, 2020 at 1:55 PM Marc Smith <msmith626@gmail.com> wrote:
>
> Hi,
>
> Apologies in advance, as I'm sure this question has been asked many
> times and there is a standard answer, but I can't seem to find it on
> forums or this mailing list.
>
> I've always observed this behavior using 'iostat', when looking at
> READ throughput numbers, the value is about 4 times more than the real
> throughput number. Knowing this, I typically look at the member
> devices to determine what throughput is actually being achieved (or
> from the application driving the I/O).
>
> Looking at the sectors read field in the 'stat' file for an MD array
> block device:
> # cat /sys/block/md127/stat && sleep 1 && cat /sys/block/md127/stat
> 93591416        0 55082801792        0       93        0        0
>   0        0        0        0        0        0        0        0
> 93608938        0 55092996456        0       93        0        0
>   0        0        0        0        0        0        0        0
>
> 55092996456 - 55082801792 = 10194664
> 10194664 * 512 = 5219667968
> 5219667968 / 1024 / 1024 = 4977
>
> This device definitely isn't doing 4,977 MiB/s. So now my curiosity is
> getting to me: Is this just known/expected behavior for the MD array
> block devices? The numbers for WRITE sectors is always accurate as far
> as I can tell. Or something configured strangely on my systems?
>
> I'm using vanilla Linux 5.4.12.

Thanks for the report. Could you please share output of

   mdadm --detial /dev/md127

and

   cat /proc/mdstat

Song

^ permalink raw reply	[flat|nested] 4+ messages in thread
* Re: MD Array 'stat' File - Sectors Read
  2020-04-09  7:11 ` Song Liu
@ 2020-04-11  4:59   ` Marc Smith
  2020-06-10 20:50     ` Marc Smith
  0 siblings, 1 reply; 4+ messages in thread
From: Marc Smith @ 2020-04-11  4:59 UTC (permalink / raw)
  To: Song Liu; +Cc: linux-raid

On Thu, Apr 9, 2020 at 3:11 AM Song Liu <song@kernel.org> wrote:
>
> On Mon, Mar 30, 2020 at 1:55 PM Marc Smith <msmith626@gmail.com> wrote:
> >
> > Hi,
> >
> > Apologies in advance, as I'm sure this question has been asked many
> > times and there is a standard answer, but I can't seem to find it on
> > forums or this mailing list.
> >
> > I've always observed this behavior using 'iostat', when looking at
> > READ throughput numbers, the value is about 4 times more than the real
> > throughput number. Knowing this, I typically look at the member
> > devices to determine what throughput is actually being achieved (or
> > from the application driving the I/O).
> >
> > Looking at the sectors read field in the 'stat' file for an MD array
> > block device:
> > # cat /sys/block/md127/stat && sleep 1 && cat /sys/block/md127/stat
> > 93591416        0 55082801792        0       93        0        0
> >   0        0        0        0        0        0        0        0
> > 93608938        0 55092996456        0       93        0        0
> >   0        0        0        0        0        0        0        0
> >
> > 55092996456 - 55082801792 = 10194664
> > 10194664 * 512 = 5219667968
> > 5219667968 / 1024 / 1024 = 4977
> >
> > This device definitely isn't doing 4,977 MiB/s. So now my curiosity is
> > getting to me: Is this just known/expected behavior for the MD array
> > block devices? The numbers for WRITE sectors is always accurate as far
> > as I can tell. Or something configured strangely on my systems?
> >
> > I'm using vanilla Linux 5.4.12.
>
> Thanks for the report. Could you please share output of
>
>    mdadm --detial /dev/md127
>

# mdadm --detail /dev/md127
/dev/md127:
           Version : 1.2
     Creation Time : Tue Mar 17 17:23:00 2020
        Raid Level : raid6
        Array Size : 17580320640 (16765.90 GiB 18002.25 GB)
     Used Dev Size : 1758032064 (1676.59 GiB 1800.22 GB)
      Raid Devices : 12
     Total Devices : 12
       Persistence : Superblock is persistent

       Update Time : Thu Apr  9 13:07:12 2020
             State : clean
    Active Devices : 12
   Working Devices : 12
    Failed Devices : 0
     Spare Devices : 0

            Layout : left-symmetric
        Chunk Size : 64K

Consistency Policy : resync

              Name : node-126c4f-1:P2024_126c4f_01  (local to host
node-126c4f-1)
              UUID : ceccb91b:1e975007:3efb5a9d:eda08d04
            Events : 79

    Number   Major   Minor   RaidDevice State
       0       8        0        0      active sync   /dev/sda
       1       8       16        1      active sync   /dev/sdb
       2       8       32        2      active sync   /dev/sdc
       3       8       48        3      active sync   /dev/sdd
       4       8       64        4      active sync   /dev/sde
       5       8       80        5      active sync
       6       8       96        6      active sync   /dev/sdg
       7       8      112        7      active sync   /dev/sdh
       8       8      128        8      active sync   /dev/sdi
       9       8      144        9      active sync   /dev/sdj
      10       8      160       10      active sync   /dev/sdk
      11       8      176       11      active sync   /dev/sdl


> and
>
>    cat /proc/mdstat

# cat /proc/mdstat
Personalities : [linear] [raid0] [raid1] [raid10] [raid6] [raid5] [raid4]
md127 : active raid6 sda[0] sdl[11] sdk[10] sdj[9] sdi[8] sdh[7]
sdg[6] sdf[5] sde[4] sdd[3] sdc[2] sdb[1]
      17580320640 blocks super 1.2 level 6, 64k chunk, algorithm 2
[12/12] [UUUUUUUUUUUU]

unused devices: <none>


Thanks; please let me know if there is any more detail I can provide.

--Marc


>
> Song

^ permalink raw reply	[flat|nested] 4+ messages in thread
* Re: MD Array 'stat' File - Sectors Read
  2020-04-11  4:59   ` Marc Smith
@ 2020-06-10 20:50     ` Marc Smith
  0 siblings, 0 replies; 4+ messages in thread
From: Marc Smith @ 2020-06-10 20:50 UTC (permalink / raw)
  To: Song Liu; +Cc: linux-raid

On Sat, Apr 11, 2020 at 12:59 AM Marc Smith <msmith626@gmail.com> wrote:
>
> On Thu, Apr 9, 2020 at 3:11 AM Song Liu <song@kernel.org> wrote:
> >
> > On Mon, Mar 30, 2020 at 1:55 PM Marc Smith <msmith626@gmail.com> wrote:
> > >
> > > Hi,
> > >
> > > Apologies in advance, as I'm sure this question has been asked many
> > > times and there is a standard answer, but I can't seem to find it on
> > > forums or this mailing list.
> > >
> > > I've always observed this behavior using 'iostat', when looking at
> > > READ throughput numbers, the value is about 4 times more than the real
> > > throughput number. Knowing this, I typically look at the member
> > > devices to determine what throughput is actually being achieved (or
> > > from the application driving the I/O).
> > >
> > > Looking at the sectors read field in the 'stat' file for an MD array
> > > block device:
> > > # cat /sys/block/md127/stat && sleep 1 && cat /sys/block/md127/stat
> > > 93591416        0 55082801792        0       93        0        0
> > >   0        0        0        0        0        0        0        0
> > > 93608938        0 55092996456        0       93        0        0
> > >   0        0        0        0        0        0        0        0
> > >
> > > 55092996456 - 55082801792 = 10194664
> > > 10194664 * 512 = 5219667968
> > > 5219667968 / 1024 / 1024 = 4977
> > >
> > > This device definitely isn't doing 4,977 MiB/s. So now my curiosity is
> > > getting to me: Is this just known/expected behavior for the MD array
> > > block devices? The numbers for WRITE sectors is always accurate as far
> > > as I can tell. Or something configured strangely on my systems?
> > >
> > > I'm using vanilla Linux 5.4.12.
> >
> > Thanks for the report. Could you please share output of
> >
> >    mdadm --detial /dev/md127
> >
>
> # mdadm --detail /dev/md127
> /dev/md127:
>            Version : 1.2
>      Creation Time : Tue Mar 17 17:23:00 2020
>         Raid Level : raid6
>         Array Size : 17580320640 (16765.90 GiB 18002.25 GB)
>      Used Dev Size : 1758032064 (1676.59 GiB 1800.22 GB)
>       Raid Devices : 12
>      Total Devices : 12
>        Persistence : Superblock is persistent
>
>        Update Time : Thu Apr  9 13:07:12 2020
>              State : clean
>     Active Devices : 12
>    Working Devices : 12
>     Failed Devices : 0
>      Spare Devices : 0
>
>             Layout : left-symmetric
>         Chunk Size : 64K
>
> Consistency Policy : resync
>
>               Name : node-126c4f-1:P2024_126c4f_01  (local to host
> node-126c4f-1)
>               UUID : ceccb91b:1e975007:3efb5a9d:eda08d04
>             Events : 79
>
>     Number   Major   Minor   RaidDevice State
>        0       8        0        0      active sync   /dev/sda
>        1       8       16        1      active sync   /dev/sdb
>        2       8       32        2      active sync   /dev/sdc
>        3       8       48        3      active sync   /dev/sdd
>        4       8       64        4      active sync   /dev/sde
>        5       8       80        5      active sync
>        6       8       96        6      active sync   /dev/sdg
>        7       8      112        7      active sync   /dev/sdh
>        8       8      128        8      active sync   /dev/sdi
>        9       8      144        9      active sync   /dev/sdj
>       10       8      160       10      active sync   /dev/sdk
>       11       8      176       11      active sync   /dev/sdl
>
>
> > and
> >
> >    cat /proc/mdstat
>
> # cat /proc/mdstat
> Personalities : [linear] [raid0] [raid1] [raid10] [raid6] [raid5] [raid4]
> md127 : active raid6 sda[0] sdl[11] sdk[10] sdj[9] sdi[8] sdh[7]
> sdg[6] sdf[5] sde[4] sdd[3] sdc[2] sdb[1]
>       17580320640 blocks super 1.2 level 6, 64k chunk, algorithm 2
> [12/12] [UUUUUUUUUUUU]
>
> unused devices: <none>
>
>
> Thanks; please let me know if there is any more detail I can provide.

I was about to follow-up on this issue, but then I noticed a couple
recent patches are being discussed and it sounds like these will
resolve what I reported above:
https://marc.info/?l=linux-raid&m=159102814820539
https://marc.info/?l=linux-raid&m=159149103212326

I'll see how these play out and report back if needed.


Thanks,

Marc



>
> --Marc
>
>
> >
> > Song

^ permalink raw reply	[flat|nested] 4+ messages in thread
end of thread, other threads:[~2020-06-10 20:50 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-03-30 20:54 MD Array 'stat' File - Sectors Read Marc Smith
2020-04-09  7:11 ` Song Liu
2020-04-11  4:59   ` Marc Smith
2020-06-10 20:50     ` Marc Smith

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.