Description of problem: The command vmstat shows the system i/o is intermittent when a command of "dd if=/dev/zero of=/file bs=1024k count=100000" is running on a cciss disk. Please look at the volume of â€bo“, you can find most of value is zero, It looks like the IO was always blocked. procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu---- r b swpd free buff cache si so bi bo in cs us sy id wa 0 5 176 14912 12224 1799968 0 0 27 6670 179 52 0 4 58 38 0 5 176 17344 12576 1801680 0 0 0 0 4538 240 0 7 25 68 1 2 176 13792 12752 1741648 0 0 84 734108 4451 176 0 21 31 47 0 4 176 16096 12832 1737440 0 0 4 12480 4711 184 0 2 50 48 0 4 176 18848 12832 1737440 0 0 32 0 4424 33 0 0 50 50 0 4 176 21152 12848 1737424 0 0 0 4 4432 169 0 0 50 50 0 3 176 26976 12848 1737424 0 0 0 0 4460 41 0 1 50 49 0 3 176 29216 12848 1737424 0 0 80 0 4479 167 0 0 50 50 0 3 176 32288 12848 1737424 0 0 112 0 4375 35 0 0 50 50 0 3 176 35488 12848 1737424 0 0 0 0 4479 167 0 1 50 49 0 3 176 38816 12848 1739488 0 0 8 0 4403 31 0 1 50 49 0 3 176 41888 12848 1739488 0 0 0 0 4464 167 0 1 50 49 0 3 176 45536 12848 1739488 0 0 0 0 4446 29 0 1 50 49 0 2 176 47840 12848 1739488 0 0 0 0 4451 171 0 1 50 49 0 2 176 50720 12848 1739488 0 0 0 0 4373 31 0 0 50 50 0 2 176 53152 12848 1739488 0 0 0 0 4426 165 0 0 50 50 0 2 176 56096 12848 1739488 0 0 0 0 4374 29 0 0 50 50 0 2 176 58080 12848 1739488 0 0 0 0 4439 165 0 0 50 50 0 2 176 61152 12848 1739488 0 0 0 0 4348 33 0 1 50 49 0 2 176 64608 12848 1739488 0 0 0 0 4439 167 0 1 50 49 0 2 176 67104 12848 1739488 0 0 0 0 4361 29 0 1 50 49 0 2 176 70304 12848 1739488 0 0 0 0 4448 165 0 0 50 50 0 2 176 72992 12848 1739488 0 0 0 0 4366 29 0 1 50 49 0 2 176 75616 12848 1739488 0 0 0 0 4431 169 0 0 50 50 1 3 176 77312 12960 1762080 0 0 8 4016 4320 31 0 15 56 29 2 2 176 12224 13408 1829744 0 0 36 58512 4836 606 0 17 33 49 0 5 176 12672 13328 1813312 0 0 4 193108 4550 97 0 14 25 61 0 5 176 17408 13360 1813280 0 0 20 0 4583 225 0 9 25 66 1 4 176 12032 13392 1821504 0 0 4 0 4473 62 0 3 23 74 0 5 176 19904 13312 1817456 0 0 4 0 4565 222 0 7 0 93 1 5 176 17312 13712 1817056 0 0 20 24372 4525 298 0 18 10 72 0 4 176 17184 14400 1812240 0 0 8 42784 4649 451 0 13 40 48 0 3 176 17184 14416 1812224 0 0 36 33828 4446 138 0 1 51 49 0 6 176 16288 14640 1760400 0 0 0 647860 4705 622 0 22 23 55 0 6 176 15136 14864 1764304 0 0 4 0 4484 407 0 5 25 70 0 6 176 18144 14896 1764272 0 0 8 0 4562 241 0 1 25 74 1 6 176 11744 15248 1776304 0 0 4 0 4476 515 0 7 25 68 0 6 176 18080 15408 1772016 0 0 8 0 4542 785 0 3 25 72 0 6 176 21856 15408 1772016 0 0 16 0 4478 78 0 1 25 74 0 6 176 17376 15760 1779920 0 0 48 0 4566 564 0 8 25 67 0 6 176 16160 15888 1783920 0 0 4 0 4459 428 0 3 25 72 0 6 176 20064 15888 1783920 0 0 80 0 4551 208 0 1 25 74 2 5 176 15072 15968 1792096 0 0 16 0 4458 78 0 2 25 73 0 6 176 14176 16336 1791728 0 0 4 0 4546 443 0 9 25 66 1 5 176 16800 16352 1793776 0 0 16 0 4456 80 0 2 25 73 0 6 176 13536 14688 1801632 0 0 0 0 4498 530 0 6 25 69 0 6 176 19168 14704 1799552 0 0 4 0 4442 97 0 2 25 73 0 6 176 22048 14704 1799552 0 0 80 0 4471 207 0 0 25 75 1 7 176 14304 14688 1783056 0 0 24 249076 4458 159 0 13 26 60 0 3 176 13152 14448 1744080 0 0 4 500260 4601 226 0 15 30 55 0 4 176 16928 14480 1739920 0 0 4 4112 4749 33 0 1 50 49 0 4 176 20192 14480 1739920 0 0 32 0 4569 167 0 1 50 49 0 4 176 22752 14496 1739904 0 0 16 4 4389 33 0 0 50 50 0 3 176 25184 14496 1739904 0 0 0 0 4422 173 0 1 50 49 0 3 176 27680 14496 1739904 0 0 16 0 4339 33 0 0 50 50 0 3 176 29984 14496 1739904 0 0 0 0 4413 165 0 0 50 50 Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. Find a mechine with HP SMART RAID CONTROLLER (any of P400, P600, E200) 2. Execute dd commands. 3. Run vmstat in another console and you can get the result. Actual results: But If we install the kernel of 2.6.9-59(RHEL4U6beta), the vmstat can report consecutive I/O. Expected results: Additional info: What the customer did: Installed the kernel of RHEL4.4 RHEL4.5. Update cciss driver to latest version on RHEL4.3 The question is: Was the output of vmstat OK or not? Is it a issue on kernel, driver or hardware? Any advice if we want to make further detail analysis. Thanks
Created attachment 221041 [details] vmstat result
What happens if you dd to the raw device, rather than through the filesystem? This would more directly test cciss, with the filesystem out of the way.
I changed the dirty_writeback_centisecs to zero, and it looks ok now. Thank you. BTW. dd to raw disk seems ok. But RHEL4u6 runs "dd file" well to both raw disk and ext3 filesystem.
per comment#3, close as "WORKSFORME" via tunable parameters: dirty_writeback_centisecs.