Re: [PATCH] Adds check for numberio during verify phase.

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

 



On Fri, Jan 24 2014, Grant Grundler wrote:
> On Fri, Jan 24, 2014 at 12:07 PM, Jens Axboe <axboe@xxxxxxxxx> wrote:
> > On Fri, Jan 24 2014, Grant Grundler wrote:
> >> [dropping jcasse since this account was deleted after his internship ended]
> >>
> >> On Fri, Jan 24, 2014 at 10:58 AM, Jens Axboe <axboe@xxxxxxxxx> wrote:
> >> > On Fri, Jan 24 2014, Grant Grundler wrote:
> >> >> Jens, Ping?
> >> >> You think you can still integrate the three patches from Juan?
> >> >
> >> > I think that would be manageable. But really a new feature (or feature
> >> > modification) like this should be accompanies by a job file example for
> >> > it. Care to provide one?
> >>
> >> Yes. Do you mind cloning a git repo?
> >
> > It was big :-)
> 
> Sorry...but I don't know how to check out a partial repo /o\
> 
> upside is you can take a look at all the fio job and autotest control
> files we are using. :)
> 
> Gwendal is cleaning up our autotest so we only use fio-2.1.2 with
> verify/integrity patches applied.
> CL is pending for that:
>     https://chromium-review.googlesource.com/#/c/183364/
> 
> We are trying to make it easier for vendors to pick up these tests and run them.
> In particular the "control.hwqual" autotest file.
> 
> ...
> >> BTW, Verification is failing on the 1m_stress control file...working
> >> on that now:
> >>     https://code.google.com/p/chromium/issues/detail?id=337651
> >>
> >> I suspect it's a problem of the control file though since we are
> >> getting this warning:
> >>     "Multiple writers may overwrite blocks that belong to other jobs.
> >> This can cause verification failures."
> >
> > Yes, with 8 jobs going, they are going to be stomping on each others
> > blocks potentially.
> 
> Yeah - that was my guess too - which means the warning is helpful.
> 
> Just to confirm: with numjobs=1, verify completes successfully.
> 
> > I queued up the 3 patches,
> 
> Awesome - thanks! :)
> 
> > but I killed the --verify-only command line
> > switch. Seems unneeded, might as well just use the job option for that.
> 
> Please reconsider. We currently use --verify.  See hardware_StorageFio.py:
> 
> hardware_StorageFio/hardware_StorageFio.py:
> ('8k_async_randwrite', ['--verifyonly'])
> 
> I want to re-use the same job file to describe the workload but
> override the "write" stage to not be executed.  Just perform verify. I
> don't care what the option is called as long as I can reuse the fio
> job file.
> 
> Having to clone a job file and make sure both files specify the same
> things is possible but provides the opportunity for simple, stupid
> mistakes. Adding --verify option eliminates that opportunity and means
> we have one less fio job file to maintain....note we have quite a few
> already.

Why not just use an environment variable, like you do for other things?
Then just have:

verify_only=${VERIFY_ONLY}

and you could easily reuse the same job file.

-- 
Jens Axboe

--
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