Flexible I/O Tester development
 help / color / mirror / Atom feed
From: Jens Axboe <axboe@kernel.dk>
To: Srinivasa Chamarthy <chamarthy.raju@gmail.com>, fio@vger.kernel.org
Subject: Re: create_on_open fails with "err=5/file:backend.c:498"
Date: Thu, 4 Aug 2016 19:43:04 -0600	[thread overview]
Message-ID: <d7afaf68-fe78-bc57-50a7-d8ddb6a77cf9@kernel.dk> (raw)
In-Reply-To: <86e61b86-3cd8-ee34-31b7-42a4b7254212@kernel.dk>

On 08/04/2016 07:39 PM, Jens Axboe wrote:
> On 08/04/2016 04:53 AM, Srinivasa Chamarthy wrote:
>> Tried the latest git, i do not get the failure now. But i have the
>> following questions
>>
>> 1) The initial failure was due to the fact that there were no files in
>> the directory to read. In that case, it may be expected that we write
>> the files first and then do read (workload defaults to read if --rw is
>> not specified).
>
> Right, but that's similar to how a normal read job without
> create_on_open would also layout the files to be able to run the read
> workload.
>
>> 2) For Write or mixed jobs, create_on_open is expected not to layout
>> the IO files before the workload. Currently with the latest git, i see
>> that it says that files are layed out (but the size is 0) before the
>> workload starts. If i do not specify create_on_open, its not laying
>> out files at all. And i have to specify create_on_open=0 explicitly to
>> layout the files completely.
>>
>> With create_on_open=1
>> # fio --name=test --directory=/testdata --nrfiles=10 --size=1g
>> --refill_buffers=1 --direct=1 --create_on_open=1 --rw=write
>> test: (g=0): rw=write, bs=4K-4K/4K-4K/4K-4K, ioengine=psync, iodepth=1
>> fio-2.13-38-gff56
>> Starting 1 process
>> test: Laying out IO file(s) (10 file(s) / 0MB)
>> Jobs: 1 (f=10): [W(1)] [100.0% done] [0KB/5684KB/0KB /s] [0/1421/0
>> iops] [eta 00m:00s]
>>
>> Without create_on_open (No Initial layout of files)
>> # fio --name=test --directory=/testdata --nrfiles=10 --size=1g
>> --refill_buffers=1 --direct=1 --rw=write
>> test: (g=0): rw=write, bs=4K-4K/4K-4K/4K-4K, ioengine=psync, iodepth=1
>> fio-2.13-38-gff56
>> Starting 1 process
>> Jobs: 1 (f=10): [W(1)] [8.2% done] [0KB/6612KB/0KB /s] [0/1653/0 iops]
>> [eta 02m:37s]
>>
>> With create_on_open=0 (its laying out files)
>> # fio --name=test --directory=/testdata --nrfiles=10 --size=1g
>> --refill_buffers=1 --direct=1 --rw=write --create_on_open=0
>> test: (g=0): rw=write, bs=4K-4K/4K-4K/4K-4K, ioengine=psync, iodepth=1
>> fio-2.13-38-gff56
>> Starting 1 process
>> test: Laying out IO file(s) (10 file(s) / 1023MB)
>> Jobs: 1 (f=10): [W(1)] [5.1% done] [0KB/6328KB/0KB /s] [0/1582/0 iops]
>> [eta 02m:49s]
>>
>> Just wanted to get confirmation on what's the default behavior. I was
>> under the impression that if create_on_open is not specified, default
>> behavior is to layout the files for jobs.
>
> Right, that was essentially the behaviour before the recent change. With
> the recent change, it'll lay them out if it needs to. It needs to, if
> the actual job doesn't do it, or if overwrite=1 is set, for instance,
> and it's a writable job.

I have reverted the change, I think the previous behavior made more
sense. I think you alluded to that as well.

-- 
Jens Axboe



  reply	other threads:[~2016-08-05  1:43 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-08-02 14:29 create_on_open fails with "err=5/file:backend.c:498" Srinivasa Chamarthy
2016-08-03 15:42 ` Jens Axboe
2016-08-03 15:54   ` Jens Axboe
2016-08-03 16:01     ` Srinivasa Chamarthy
2016-08-04 10:53       ` Srinivasa Chamarthy
2016-08-05  1:39         ` Jens Axboe
2016-08-05  1:43           ` Jens Axboe [this message]
2016-08-05  1:43           ` Srinivasa Chamarthy

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=d7afaf68-fe78-bc57-50a7-d8ddb6a77cf9@kernel.dk \
    --to=axboe@kernel.dk \
    --cc=chamarthy.raju@gmail.com \
    --cc=fio@vger.kernel.org \
    /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