On 6 January 2018 at 14:35, Gnana Sekhar <kgsgnana2020@xxxxxxxxx> wrote: > > If I wanted the pattern written to be random, is it correct to specify > option -random_generator. Probably not - random_generator only controls which algorithm is used to generate random numbers so in this case it's not the option you're after. > > If random_generator is the option to use which option should I specify If you don't enable a fixed pattern and you enable verification at write time then the data fio writes into each block will be random bar the verification header. $ rm -f /tmp/fio.tmp $ fio --rw=write --filename=/tmp/fio.tmp --bs=64 --size=512 --verify=crc32c --name=write_then_verify [...] $ hexdump -C /tmp/fio.tmp 00000000 ca ac 05 00 40 00 00 00 3c 54 09 c5 01 cc a4 00 |....@...<T......| 00000010 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 |................| 00000020 01 00 00 00 d5 4b 36 f6 4b ae d6 e9 6c c7 4e 18 |.....K6.K...l.N.| 00000030 ce d0 f8 3c 26 d8 f1 17 19 1a 38 5f 88 1c 0d 11 |...<&.....8_....| 00000040 ca ac 05 00 40 00 00 00 2e 56 db f6 8c 0a bb 88 |....@....V......| 00000050 40 00 00 00 00 00 00 00 0a 42 05 00 b2 44 0d 00 |@........B...D..| 00000060 01 00 01 00 1a d5 68 80 17 a8 c3 55 3b 05 d3 1f |......h....U;...| 00000070 60 95 97 0f 3d cc 06 05 ac f2 42 d7 bb 51 22 18 |`...=.....B..Q".| 00000080 The above shows that when you get past the verification header (which is at the beginning of the blocks and in this case is 44 bytes big) the rest of the 128 byte blocks are different to each other. Note: the random sequence that is used is based off a seed controlled by randseed (http://fio.readthedocs.io/en/latest/fio_doc.html#cmdoption-arg-randseed ). By default, fio uses the same seed each time you run it but this is controllable - see randrepeat (http://fio.readthedocs.io/en/latest/fio_doc.html#cmdoption-arg-randrepeat ) and related options. > during verify. Will FIO keep track of random patterns written to verify Sort of yes but what it uses to do so depends on how and when you choose to do the verification. If you are happy to do writes and let fio immediately do a verification pass within the same job then you will fine with the example above (also see https://github.com/axboe/fio/blob/master/examples/basic-verify.fio ). If you want to split things into entirely separate jobs then the second job will have to be a read job (or a verify_only write job). fio will entirely determine a block's correctness by using the data in that block's verification header: $ ./fio --rw=write --filename=/tmp/fio.tmp --bs=64 --size=128 --verify=crc32c --do_verify=0 --name=write_with_verify_headers $ ./fio --rw=read --filename=/tmp/fio.tmp --bs=64 --size=128 --verify=crc32c --name=verify $ # Following generates a spurious error about no I/O but actually works fine $ ./fio --rw=write --filename=/tmp/fio.tmp --bs=64 --size=128 --verify=crc32c --verify_only=1 --name=verify $ # Damage the end of the second block so we can show verification failure $ dd if=/dev/zero of=/tmp/fio.tmp bs=1 seek=127 count=1 $ ./fio --rw=read --filename=/tmp/fio.tmp --bs=64 --size=128 --verify=crc32c --name=verify_failure See the verification section of the documentation (e.g. http://fio.readthedocs.io/en/latest/fio_doc.html#verification ) for more verification details and options. -- Sitsofe | http://sucs.org/~sits/ -- 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
