> > fio --name=randwrite --ioengine=libaio --iodepth=64 --rw=randrw --rwmixread=50 --bs=4k -- > direct=1 -filename=data --numjobs=1 --runtime 36000 --verify=md5 --verify_async=4 -- > verify_backlog=100000 --verify_dump=1 --verify_fatal=1 --time_based --group_reporting -- > continue_on_error = io > > If an error happens on an IO then how verification happens of that IO. Does FIO completely skip > that IO for any future verification No, it doesn't do anything special to remember them. Something paralleling the randommap feature could be used to knock out bad blocks, if you need such an option. If using fio to stress test error handling, you might want it to keep trying bad blocks. Errors might go away on their own, and are usually fixed by overwriting the block, so giving up on a block forever after the first error wouldn't simulate real program behavior. > or does it maintain multiple checksums and passes if it verifies > with any of the previous checksums. No; the number of possible combinations of good + bad blocks would be huge.
