From: Jens Axboe <axboe@kernel.dk>
To: Ross Becker <ross.becker@gmail.com>
Cc: fio@vger.kernel.org
Subject: Re: Fio issues with Solaris /dev/rdsk devices
Date: Tue, 22 Nov 2011 09:55:37 +0100 [thread overview]
Message-ID: <4ECB6389.3040003@kernel.dk> (raw)
In-Reply-To: <CAEAE17F.578FF%ross.becker@gmail.com>
On 2011-11-18 00:06, Ross Becker wrote:
> I am having some difficulty with fio using /dev/rdsk devices under
> Solaris. My best guess at this point is that fio is doing something
> incorrect when it attempts to determine the size of a /dev/rdsk device.
> Everything works if I use /dev/rdsk/c0t5SOMEDISKd0p0, but if I create a
> partition which is not as large as the complete disk, when I point fio at
> the /dev/rdsk device (e.g. /dev/rdsk/c0t5SOMEDISKd0p1), fio promptly bombs
> out. If I point fio at the /dev/dsk device (e.g.
> /dev/dsk/c0t5SOMEDISKd0p1), fio works, though I typically get half to a
> third of the performance using the /dev/dsk devices.
>
> I have tested using the "dd" utility, and validated that "dd" will use the
> /dev/rdsk devices and access only part of the drive without a problem.
>
> I am testing on solaris 11 x86, and I have tested fio versions 1.54 and
> 1.99. Both versions of fio failed the same way.
>
> Here's what fio looks like when it bombs out:
> rand-read: (g=0): rw=randrw, bs=32K-32K/32K-32K, ioengine=solarisaio,
> iodepth=8
> ...
> rand-read: (g=0): rw=randrw, bs=32K-32K/32K-32K, ioengine=solarisaio,
> iodepth=8
> fio 1.54
> Starting 4 processes
> fio: io_u error on file /dev/rdsk/c0t500151795968D5ABd0p1: I/O error
> read offset=134944555008, buflen=32768
> fio: io_u error on file /dev/rdsk/c0t500151795968D5ABd0p1: I/O error
> write offset=134455656448, buflen=32768
> fio: pid=1382, err=5/file:io_u.c:1197, func=io_u error, error=I/O error
> fio: io_u error on file /dev/rdsk/c0t500151795968D5ABd0p1: I/O error
> read offset=156234678272, buflen=32768
> fio: pid=1383, err=5/file:io_u.c:1197, func=io_u error, error=I/O error
> fio: io_u error on file /dev/rdsk/c0t500151795968D5ABd0p1: I/O error
> read offset=131076620288, buflen=32768
> fio: io_u error on file /dev/rdsk/c0t500151795968D5ABd0p1: I/O error
> write offset=155262877696, buflen=32768
> fio: pid=1384, err=5/file:io_u.c:1197, func=io_u error, error=I/O error
> fio: io_u error on file /dev/rdsk/c0t500151795968D5ABd0p1: I/O error
> read offset=156624125952, buflen=32768
> fio: pid=1385, err=5/file:io_u.c:1197, func=io_u error, error=I/O error
>
>
> Here is my fio config file:
>
> [global]
> bs=32k
> ioengine=solarisaio
> iodepth=8
> openfiles=1024
> runtime=60
> ramp_time=10
> filename=/dev/rdsk/c0t500151795968D5ABd0p1
>
>
>
> [rand-read]
> rw=randrw
> numjobs=4
> rwmixwrite=20
> norandommap=1
> file_service_type=random
> group_reporting
>
>
>
> What additional information, if any should I provide?
I think you provided all you need. When I get my hands on a Solaris box,
I can take a look. Sounds like the ioctl for retrieving device size
always returns the full size, not just the partition. So should be en
easy fix, perhaps there's an alternative API that could be used.
--
Jens Axboe
prev parent reply other threads:[~2011-11-22 8:55 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-11-17 23:06 Fio issues with Solaris /dev/rdsk devices Ross Becker
2011-11-22 8:55 ` Jens Axboe [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4ECB6389.3040003@kernel.dk \
--to=axboe@kernel.dk \
--cc=fio@vger.kernel.org \
--cc=ross.becker@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox