Hi, On Sun, 1 Dec 2019 at 13:16, Roy Shterman <roys@xxxxxxxxxxxxxxxxx> wrote: > > Hi, > > I'm running fio job with trigger, and using load/save_verify_state to > run verify phase afterwards. > Both write and read have the same offsets and sizes and both are using > non random patterns. > From time to time I see that the fio read verify actually reads > offsets bigger than the latest offset in the file saved during write > phase. > > Both workloads are running with iodepth of 300, I wonder if the > iodepth affect somehow the end offset during read phase? > > Could it be that fio during read is reading from last_sector_written + iodepth? Do you have a simple job that can demonstrate this issue? A depth of 300 is extremely high so anything you can do to demonstrate the issue with smaller depths would be great! -- Sitsofe | http://sucs.org/~sits/
