On Wed, Aug 14, 2019 at 2:38 PM Olivier AUDRY wrote:
> let's test random write
> rbd -p kube bench kube/bench --io-type write --io-size 8192 --io-threads 256
> --io-total 10G --io-pattern rand
> elapsed: 125 ops: 1310720 ops/sec: 10416.31 bytes/sec: 85330446.58
>
> dd if=/dev/zero of=test b
On Wed, Aug 14, 2019 at 2:49 PM Paul Emmerich wrote:
>
> On Wed, Aug 14, 2019 at 2:38 PM Olivier AUDRY wrote:
> > let's test random write
> > rbd -p kube bench kube/bench --io-type write --io-size 8192 --io-threads
> > 256 --io-total 10G --io-pattern rand
> > elapsed: 125 ops: 1310720 ops/s
hello
I mean a filesystem mounted on top of a mapped rbd
rbd create --size=10G kube/benchrbd feature disable kube/bench object-
map fast-diff deep-flattenrbd map bench --pool kube --name
client.admin/sbin/mkfs.ext4 /dev/rbd/kube/bench mount
/dev/rbd/kube/bench /mnt/cd /mnt/
about the bench I did
rbd -p kube bench kube/bench --io-type write --io-threads 1 --io-total
10G --io-pattern rand
elapsed:14 ops: 262144 ops/sec: 17818.16 bytes/sec: 72983201.32
It's a totally unreal number. Something is wrong with the test.
Test it with `fio` please:
fio -ioengine=rbd -name=test -bs=4k
hello
here the result :
fio --ioengine=rbd --name=test --bs=4k --iodepth=1 --rw=randwrite --
runtime=60 -pool=kube -rbdname=bench
test: (g=0): rw=randwrite, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T)
4096B-4096B, ioengine=rbd, iodepth=1
fio-3.12
Starting 1 process
Jobs: 1 (f=1): [w(1)][100.0%][w=2
hello
just for the record the nvme disk are pretty fast.
dd if=/dev/zero of=test bs=8192k count=100 oflag=direct
100+0 records in
100+0 records out
838860800 bytes (839 MB, 800 MiB) copied, 0.49474 s, 1.7 GB/s
oau
Le vendredi 16 août 2019 à 13:31 +0200, Olivier AUDRY a écrit :
> hello
>
> here
Now to go for "apples to apples" either run
fio -ioengine=libaio -name=test -bs=4k -iodepth=1 -direct=1 -fsync=1
-rw=randwrite -runtime=60 -filename=/dev/nvmeX
to compare with the single-threaded RBD random write result (the test is
destructive, so use a separate partition without dat
hello
here on the nvme partition directly
- libaio randwrite /dev/nvme1n1p4 => WRITE: bw=12.1MiB/s (12.7MB/s),
12.1MiB/s-12.1MiB/s (12.7MB/s-12.7MB/s), io=728MiB (763MB), run=60001-
60001msec
- libaio randread /dev/nvme1n1p4 => READ: bw=35.6MiB/s (37.3MB/s),
35.6MiB/s-35.6MiB/s (37.3MB/s-37.3MB/
Ok I read your link. My ssds are bad. They got capacitors ... I don't
choose them. They come with the hardware I rent. Perhaps it will be
better to switch to hdd. I cannot even but journal on them ... bad news
:(
Le vendredi 16 août 2019 à 17:37 +0200, Olivier AUDRY a écrit :
> hello
>
> here on
- libaio randwrite
- libaio randread
- libaio randwrite on mapped rbd
- libaio randread on mapped rbd
- rbd read
- rbd write
recheck RBD with RAND READ / RAND WRITE
you're again comparing RANDOM and NON-RANDOM I/O
your SSDs aren't that bad, 3000 single-thread iops isn't the worst
possib
fio -ioengine=rbd -name=test -bs=4M -iodepth=32 -rw=randwrite
-runtime=60 -pool=kube -rbdname=bench
WRITE: bw=89.6MiB/s (93.9MB/s), 89.6MiB/s-89.6MiB/s (93.9MB/s-
93.9MB/s), io=5548MiB (5817MB), run=61935-61935msec
fio -ioengine=rbd -name=test -bs=4M -iodepth=32 -rw=randread
-runtime=60 -pool=k
And once more you're checking random I/O with 4 MB !!! block size.
Now recheck it with bs=4k.
___
ceph-users mailing list -- ceph-users@ceph.io
To unsubscribe send an email to ceph-users-le...@ceph.io
RBD
fio -ioengine=rbd -name=test -bs=4k -iodepth=32 -rw=randread
-runtime=60 -pool=kube -rbdname=bench
READ: bw=21.8MiB/s (22.8MB/s), 21.8MiB/s-21.8MiB/s (22.8MB/s-22.8MB/s),
io=1308MiB (1371MB), run=60011-60011msec
fio -ioengine=rbd -name=test -bs=4k -iodepth=32 -rw=randwrite
-runtime=60 -pool
fio -ioengine=libaio -name=test -bs=4k -iodepth=32 -rw=randread
-runtime=60 -filename=/dev/rbd/kube/bench
Now add -direct=1 because Linux async IO isn't async without O_DIRECT.
:)
+ Repeat the same for randwrite.
___
ceph-users mailing list -- ceph-
on a new ceph cluster with the same software and config (ansible) on
the old hardware. 2 replica, 1 host, 4 osd.
RBD
fio -ioengine=rbd -name=test -bs=4k -iodepth=32 -rw=randread
-runtime=60 -pool=kube -rbdname=bench
READ: bw=120MiB/s (126MB/s), 120MiB/s-120MiB/s (126MB/s-126MB/s),
io=7189MiB (75
on a new ceph cluster with the same software and config (ansible) on
the old hardware. 2 replica, 1 host, 4 osd.
=> New hardware : 32.6MB/s READ / 10.5MiB WRITE
=> Old hardware : 184MiB/s READ / 46.9MiB WRITE
No discussion ? I suppose I will keep the old hardware. What do you
think ? :D
In fac
Write and read with 2 hosts 4 osd :
mkfs.ext4 /dev/rbd/kube/bench
mount /dev/rbd/kube/bench /mnt/
dd if=/dev/zero of=test bs=8192k count=1000 oflag=direct
8388608000 bytes (8.4 GB, 7.8 GiB) copied, 117.541 s, 71.4 MB/s
fio -ioengine=libaio -name=test -bs=4k -iodepth=32 -rw=randwrite
-direct=1 -ru
This probably muddies the water. Note Active cluster with around 22
read/write IOPS and 200kB read/write
A CephFS mounted with 3 hosts 6 osd per host with 8G public and 10G
private networking for Ceph.
No SSDs and mostly WD Red 1T 2.5" drives some are HGST 1T 7200.
root@blade7:~# fio -ioengine=li
18 matches
Mail list logo