in example files, stonewall is usually the last option, and is usually part of all or none of the jobs. learning by example didn't work for me :) yes, "Make the job containing the stonewall parameter wait for preceding jobs in the job file to exit before starting." is clearer. may I suggest an alternative? maybe a special reserved job-name [stonewall] can make it clear, while maintaining backwards compatibility (assuming the unlikely case of existing files with such a job name). I would expect such a dummy-job to be empty and contain the implicit stonewall option, preserving existing behavior. introducing dummy jobs (e.g., rw=null) can also serve other features, such as waiting/thinking for a while between jobs (letting cache cool off) or triggering external commands. On Wed, Jul 12, 2017 at 9:19 PM, Sitsofe Wheeler <sitsofe@xxxxxxxxx> wrote: > Hi, > > Would changing the HOWTO to say: > "Make the job containing the stonewall parameter wait for preceding > jobs in the job file to exit before starting." > be any clearer? What did you find that made you expect the original > behaviour and what did you find that made you change your > understanding? > > On 12 July 2017 at 08:01, Ido Ben-Tsion <idob@xxxxxxxxxxxxx> wrote: >> ok, I understand. I didn't think of stonewall as being part of a job. >> I thought of it as a separator between jobs, unrelated to any specific >> job. which is not what the documentation explains. >> >> On Wed, Jul 12, 2017 at 9:26 AM, Sitsofe Wheeler <sitsofe@xxxxxxxxx> wrote: >>> Hi, >>> >>> On 12 July 2017 at 06:13, Ido Ben-Tsion <idob@xxxxxxxxxxxxx> wrote: >>>> fio version: fio-2.21-37-ga2c9 >>>> >>>> fio file excerpt: >>>> [global] >>>> ioengine=sg >>>> invalidate=1 >>>> ramp_time=5 >>>> iodepth=16 >>>> runtime=300 >>>> time_based >>>> bs=64k >>>> norandommap >>>> loops=5 >>>> >>>> [seq_read-2-jobs] >>>> rw=read >>>> numjobs=2 >>>> stonewall >>>> >>>> [randread] >>>> rw=randread >>>> [seq_write] >>>> rw=write >>>> stonewall >>>> ... >>>> >>>> In the report, I see jobs with groupid 0 which are not supposed to be >>>> on same group: >>>> ...,jobname,groupid,... >>>> ...,seq_read-2-jobs,0,... >>>> ...,seq_read-2-jobs,0,... >>>> ...,randread,0,... >>>> ...,seq_write,1,... >>> >>> I'm not sure I follow - which jobs are wrong? Looking at the opening >>> sentence explanation (and the alternative name) for stonewall in the >>> HOWTO (http://fio.readthedocs.io/en/latest/fio_man.html#cmdoption-arg-stonewall >>> ) seems to describe the behaviour you're seeing. >>> >>> Can you be more explicit as to the problem? > > -- > Sitsofe | http://sucs.org/~sits/ -- Ido Ben-Tsion Ackerstein Towers, A Tower, Fl.4 Hamenofim St. P.O.B 12696, Hertzelia Pituach, 46725 T +972.9.970.4000, F +972.9.970.4001 M +972.5X.XXX.XXXX www.infinidat.com -- 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
