Re: Running multiple jobs on a raw drive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi,

On 25 May 2017 at 18:15, Ratn, Prasun <Prasun.Ratn@xxxxxxx> wrote:
>
> I am trying to run fio on a raw NVMe drive using 2 jobs with QD=4.
>
> I am not sure how to specify the offsets for each job to write to. I tried using offset_increment but that doesn't seem to work.
>
> The first run with 1 job runs smoothly.
>
>         $ sudo fio --rw=randwrite --bs=4k --name=/dev/nvme0n1 --runtime=10 --direct=1 --ioengine=libaio --iodepth=4
[...]
>
> When I try specifying numjobs=2 with offset_increment, I get an error and only 1 process is launched.
>
>
>         $ sudo fio --rw=randwrite --bs=4k --name=/dev/nvme0n1 --runtime=10 --direct=1 --ioengine=libaio --iodepth=4 --numjobs=2 --offset_increment=81920m
>         /dev/nvme0n1: (g=0): rw=randwrite, bs=4k-4k/4k-4k/4k-4k, ioengine=libaio, iodepth=4
>         ...
>         fio-2.2.8
>         starting 2 processes
>         /dev/nvme0n1: you need to specify size=
>         fio: pid=0, err=22/file:filesetup.c:802, func=total_file_size, error=invalid argument
>         ^cbs: 1 (f=1): [w(1),x(1)] [54.5% done] [0kb/510.2mb/0kb /s] [0/131k/0 iops] [eta 00m:05s]
>         fio: terminating on signal 2
>
> What's the best way to have multiple jobs run on a single raw drive.

Because you don't want fio to "invent" the name of the file you are
are doing I/O to you're better off using a filename job parameter and
then the name of your job can be whatever you like e.g.

fio --rw=randwrite --bs=4k --filename=/dev/nvme0n1 --runtime=10
--direct=1 --ioengine=libaio --iodepth=4 --numjobs=2
--offset_increment=81920m --name=randomwrites

(all the job parameters were global) or

fio --rw=randwrite --bs=4k --name=randomwrites --filename=/dev/nvme0n1
--runtime=10 --direct=1 --ioengine=libaio --iodepth=4 --numjobs=2
--offset_increment=81920m

(two parameters were global but achieving an identical effect)

See http://fio.readthedocs.io/en/latest/fio_doc.html#cmdoption-arg-filename
for details and
http://fio.readthedocs.io/en/latest/fio_doc.html#cmdoption-arg-size
for how the size semantics change when filename is used. Tip: if you
don't want the regions the jobs do I/O against to overlap (which I'm
guessing is why you already have offset_increment) you'll need to set
the size parameter too.

-- 
Sitsofe | http://sucs.org/~sits/
--
To unsubscribe from this list: send the line "unsubscribe fio" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [Linux Kernel]     [Linux SCSI]     [Linux IDE]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux SCSI]

  Powered by Linux