On Mon, Dec 28, 2020 at 05:02:50PM +0800, yukuai (C) wrote: > Hi > > On 2020/12/28 16:28, Ming Lei wrote: > > Another candidate solution may be to always return true from hctx_may_queue() > > for this kind of queue because queue_depth has provided fair allocation for > > each LUN, and looks not necessary to do that again. > > If always return true from hctx_may_queue() in this case, for example, > we set queue_depth to 128(if can't, the biggger, the better) for all > disks, and test with numjobs=64. The result should be one disk with high > iops, and the rest very low. So I think it's better to ensure the max > tags a disk can get in this case. Just wondering why you try to set 128 via sysfs for all disks? If you do that, you should know the potential result given the whole tags queue depth is just 128. Thanks, Ming
