Hi Ankit, Thanks for the quick response. The latest fio master branch works as expected, however I still see more IO requests being issued. Run status group 0 (all jobs): READ: bw=1023B/s (1023B/s), 1023B/s-1023B/s (1023B/s-1023B/s), io=20.0KiB (20.5kB), run=20001-20001msec To investigate more I looked at the blktrace log we can see that fio issues more requests per second even with rate_ignore_thinktime. In the following blktrace logs, we can see that in the first io phase, 2 requests were submitted and in the second io phase, 3 IO requests were submitted. is there any other parameter I should use to prevent this? 259,3 26 2 0.000003132 2713717 Q R 1650183824 + 8 [fio] 259,3 26 5 0.999898056 2713717 Q R 1692964272 + 8 [fio] 259,3 26 8 10.000133491 2713717 Q R 1699437744 + 8 [fio] 259,3 26 11 10.999891950 2713717 Q R 1675960000 + 8 [fio] 259,3 26 14 11.999888574 2713717 Q R 1671675368 + 8 [fio] Thanks & Regards, Lokesh On Mon, Feb 27, 2023 at 9:45 AM Ankit Kumar <ankit1455@xxxxxxxxx> wrote: > > Hi Lokesh, > > Can you please try out with the latest fio master branch. > It should work as per the documentation. > > Regards > Ankit Kumar > > > On Mon, Feb 27, 2023 at 10:58 PM lokesh jaliminche > <lokesh.jaliminche@xxxxxxxxx> wrote: > > > > Hi Everyone, > > > > As mentioned in the readthedocs, thinktime, and thinktime_iotime > > can be used to simulate the IO phase followed by the idle phase > > of the workload. For example, --thinktime_iotime=1s --thinktime=9s > > repeat the 10-second cycle with IOs for 1 second and stall for > > 9 seconds. However, in my test for the first 10sec, fio keeps submitting > > without respecting thinktime and thinktime_iotime. After the first 10 > > seconds fio starts simulating mentioned IO phases according to parameters. > > Following is my fio script. > > > > [global] > > direct=1 > > runtime=20 > > time_based > > [job_1] > > rw=randread > > bs=4k > > rate_iops=1 > > numjobs=1 > > ioengine=libaio > > iodepth=1 > > offset=40G > > size=30G > > filename=/dev/nvme0n1p2 > > thinktime_iotime=1s > > thinktime=9s > > rate_ignore_thinktime=1 > > > > Results: > > Run status group 0 (all jobs): > > READ: bw=2457B/s (2457B/s), 2457B/s-2457B/s (2457B/s-2457B/s), > > io=48.0KiB (49.2kB), run=20001-20001msec > > > > From the results, we can see that the total IO is 48KiB it should be > > ideally 8KiB. > > > > Is this a known issue? > > > > Thanks & Regards, > > Lokesh
