Message-ID: <COL129-W590158F348C93F28E325FE92010@phx.gbl>
From: Steven Jones <steven.jones1201@outlook.com>
To: didier <did447@gmail.com>
CC: "pgsql-performance@postgresql.org" <pgsql-performance@postgresql.org>
Subject: Re: Very slow checkpoints
Date: Thu, 19 Mar 2015 05:28:40 +0000
Importance: Normal
In-Reply-To: 
 <CAJRYxuKZpxNJVVGBzuN8OrCDFFOybWhVb6jZQYj3zKWtj4T4Hg@mail.gmail.com>
References: <COL129-W88EAA235AC3967FF7F5D8792000@phx.gbl>,
 <CAJRYxuKZpxNJVVGBzuN8OrCDFFOybWhVb6jZQYj3zKWtj4T4Hg@mail.gmail.com>
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
Precedence: bulk
Sender: pgsql-performance-owner@postgresql.org

Hi,

>>
>> Device: rrqm/s wrqm/s r/s w/s rkB/s wkB/s avgrq-sz
>> avgqu-sz await r_await w_await svctm %util
>> sda 0.00 0.00 0.00 5.00 0.00 2056.00 822.40
>> 0.00 0.00 0.00 0.00 0.00 0.00
>> sdb 0.00 0.00 1055.00 549.00 41166.50 22840.00 79.81
>> 5.28 3.28 4.94 0.10 0.62 100.00
> Your sdb is saturated...

Yes that's what iostat seems to indicate, but it's weird because at the sam=
e time it is reporting 100% io utilization I can hit the disk write (seq) a=
t> 250Mbyte/sec:

# sync;time bash -c "(dd if=3D/dev/sda1 of=3Dbf bs=3D8k count=3D500000; syn=
c)"
500000+0 records in
500000+0 records out
4096000000 bytes (4.1 GB) copied, 14.8575 s, 276 MB/s

real	0m14.896s
user	0m0.068s
sys	0m10.157s


> Why are checkpointer process and writer process reading at 18 MB/s ?
> I have no experience with zfs but could it be related to COW and
> recordsize? I have no idea if these reads are counted in iotop output
> though.

In general, some random disk write benchmarks and varying block sizes don't=
 have a huge effect. =A0But for some reason the checkpointing process is ju=
st simply writing checkpoints too slowly. In the meantime the COPY is pilin=
g up logs while the previous checkpoint is still being written, so the next=
 one starts straight away and no setting is able to split them up.


Regards,
Steve=20=09=09=20=09=20=20=20=09=09=20=20

--=20
Sent via pgsql-performance mailing list (pgsql-performance@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-performance