Hi, I am using blktrace to understand Block Layer behavior in a multi-thread IO workload with high IO load. I can observe some duplicated LBA request in block layer. For example, I have only one write request to LBA 47568 in my original workload. However, blktrace shows 121 requests writing LBA 47568 !!! below is a portion of blkparse log which filter only I events (IO inserted onto request queue). I am trying to understand why this phenomena happen and how can I eliminate that. I picked noop IO scheduler and value 2 in /sys/block/sdX/nomerge to avoid IO merges. I do appreciate your help in advance. 8 240 0 7539 0.073300123 32282 I WS 47576 + 8 [hfplayer] 8 240 5 4462 0.073303974 32281 I WS 47568 + 8 [hfplayer] 8 240 11 4462 0.073306337 32283 I WS 47584 + 8 [hfplayer] 8 240 6 4462 0.073307970 32284 I WS 47592 + 8 [hfplayer] 8 240 10 4462 0.073309010 32285 I WS 47600 + 8 [hfplayer] 8 240 5 4466 0.073311526 32281 I WS 47568 + 8 [hfplayer] 8 240 5 4469 0.073316215 175 I WS 47568 + 8 [kworker/5:1] 8 240 7 4462 0.073352932 32286 I WS 47608 + 8 [hfplayer] 8 240 7 4466 0.073354061 32286 I WS 47568 + 8 [hfplayer] 8 240 8 4468 0.073358935 32280 I WS 47624 + 8 [hfplayer] 8 240 4 4468 0.073360550 32279 I WS 47616 + 8 [hfplayer] 8 240 8 4472 0.073362145 32280 I WS 47568 + 8 [hfplayer] 8 240 5 4473 0.073362711 32281 I WS 47632 + 8 [hfplayer] 8 240 5 4477 0.073364177 32281 I WS 47568 + 8 [hfplayer] 8 240 0 7545 0.073407427 32282 I WS 47640 + 8 [hfplayer] 8 240 0 7549 0.073409589 32282 I WS 47568 + 8 [hfplayer] -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
