Flexible I/O Tester development
 help / color / mirror / Atom feed
From: Jens Axboe <axboe@kernel.dk>
To: Thierry BERTAUD <tbertaud@conforama.fr>,
	"fio@vger.kernel.org" <fio@vger.kernel.org>
Subject: Re: fio-2c2c93d and with a lot of numjobs causing fio Jobs to Time Out
Date: Wed, 30 Dec 2015 21:02:57 -0700	[thread overview]
Message-ID: <5684A8F1.6010105@kernel.dk> (raw)
In-Reply-To: <VI1PR05MB13435F5190D9D9A17C59315BB1FD0@VI1PR05MB1343.eurprd05.prod.outlook.com>

On 12/30/2015 11:34 AM, Thierry BERTAUD wrote:
> Hello Jens,
>
>> That's expected - when the main thread goes to sleep, it checks for this
>> status dump file. If it exists, then it'll show the current summary of
>> stats, then unlink it.
> Thanks for explain this point.
>
> Just to confirm my understanding the following option with engine 'posixaio'
> iodepth=128 and numjobs=128 =>
> I don't use 'thread' option so i think with iodpeth=128 the main fio process will fork to create 128 process.
> With numjob=128 that imply, i'll clone each forked process from fio (due to iodepth) that's imply 16384 process (128 * 128).
> If i use threads enabled, with iodepth=128 the main fio process will create with pthread_create 128 threads for itself.
> With numjobs=128, the ain fio process will clone to have  16384 threads into itselft  or it will fork to create  128 process and each process has 128 threads.
>
>> That's the main thread, what are the IO threads doing?
> If i follow disk I/O there are no activities.
> This kind of information for fio process can help or not to answer the questions?
> The list of thread by process this:
>     # ps -mo THREAD
>         USER      PID     PPID        TID ST  CP PRI SC    WCHAN        F     TT BND COMMAND
>         root  9437428 10551472          - A    0  60  1        -   200001  pts/0   - ksh
>            -        -        -   18350143 S    0  60  1        -    10400      -   - -
>       padmin 10551472 11534386          - A    0  60  1        -   240001  pts/0   - -rksh
>            -        -        -   49873143 S    0  60  1        -    10400      -   - -
>         root 11206870  9437428          - A    1  60  1        -   200001  pts/0   - bash
>            -        -        -   48759011 S    1  60  1        -   410400      -   - -
>         root 62914724 11206870          - A    6  63  1        -   200001  pts/0   - ps -mo THREAD
>            -        -        -  116850921 R    6  63  1        -   400000      -   - -
>     #
> Or thread for a given process like this:
>     # ps -mp 9437428 -o THREAD
>         USER      PID     PPID        TID ST  CP PRI SC    WCHAN        F     TT BND COMMAND
>         root  9437428 10551472          - A    0  60  1        -   200001  pts/0   - ksh
>            -        -        -   18350143 S    0  60  1        -    10400      -   - -
>     #
>
> I tried to use gdb against the live process but i don't how to progress with it?
> I can do gdb against the executable also but i don't know how to progress whit it?

Against the live process it the easiest. Just do gdb -p <pid> and 
attach. If you don't use threads from fio, then they will each be a 
separate process.

I'll be taking a bit of time off, but I should have access to an AIX 
that I can test on sometime next week. What version are you on?

-- 
Jens Axboe



  reply	other threads:[~2015-12-31  4:02 UTC|newest]

Thread overview: 24+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-12-29  8:29 fio-2c2c93d and with a lot of numjobs causing fio Jobs to Time Out Thierry BERTAUD
2015-12-29 14:30 ` Thierry BERTAUD
2015-12-30 17:08 ` Jens Axboe
2015-12-30 18:34   ` Thierry BERTAUD
2015-12-31  4:02     ` Jens Axboe [this message]
2015-12-31 16:18       ` Thierry BERTAUD
2015-12-31 16:33         ` Andrey Kuzmin
2016-01-01 12:43           ` Thierry BERTAUD
2016-01-01 12:46             ` Thierry BERTAUD
2016-01-01 12:54             ` Andrey Kuzmin
2016-01-01 14:00               ` Thierry BERTAUD
2016-01-01 14:04                 ` Andrey Kuzmin
2016-01-01 14:49                   ` Thierry BERTAUD
2016-01-01 16:19                     ` Andrey Kuzmin
2016-01-01 17:31                       ` Thierry BERTAUD
2016-01-18 23:58                         ` How to stress a 8Gbps card to get 500 000 IOPS or 1600 MB/s Thierry BERTAUD
2016-01-19  0:15                           ` David Nellans
2016-01-19  0:29                             ` Thierry BERTAUD
     [not found]                               ` <CAB-428m2HJBUPdtYsCKtxfYPuscV1i=ucCvCAPh4ShqWPNmTXw@mail.gmail.com>
     [not found]                                 ` <VI1PR05MB1343CA9499CC6AB3B7A0D2C0B1C10@VI1PR05MB1343.eurprd05.prod.outlook.com>
2016-01-19  2:45                                   ` Alireza Haghdoost
2016-01-19  0:33                             ` Kulkarni, Vasu
2016-01-19  0:40                               ` David Nellans
2016-01-19  1:04                                 ` Thierry BERTAUD
2016-01-19  1:15                                   ` Alireza Haghdoost
2016-01-19  1:59                                     ` Thierry BERTAUD

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=5684A8F1.6010105@kernel.dk \
    --to=axboe@kernel.dk \
    --cc=fio@vger.kernel.org \
    --cc=tbertaud@conforama.fr \
    /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