On Thu, Mar 29, 2018 at 03:24:38PM -0400, John Pittman wrote:
> Commit ac8fa4196d20 ("md: allow resync to go faster when there is
> competing IO.") introduced performance gaining features that are
> very beneficial to solid state media. When the same features
> are used in an environment with high I/O and rotational media,
> extreme performance degradation of non-sync I/O can be seen.
> Rotational devices are still slow by comparison, so to avoid these
> performance issues, only apply the newer code path to non-rotational
> devices.
Do we know the reason why this doesn't work for rotational media? I'd rather
understand the reason and fix the real problem instead of apply an arbitrary
'fix'.
Thanks,
Shaohua
> Signed-off-by: John Pittman <jpittman@xxxxxxxxxx>
> ---
> drivers/md/md.c | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/md/md.c b/drivers/md/md.c
> index 254e44e..00dc3c4 100644
> --- a/drivers/md/md.c
> +++ b/drivers/md/md.c
> @@ -8496,11 +8496,13 @@ void md_do_sync(struct md_thread *thread)
> /((jiffies-mddev->resync_mark)/HZ +1) +1;
>
> if (currspeed > speed_min(mddev)) {
> - if (currspeed > speed_max(mddev)) {
> + if (currspeed > speed_max(mddev) || (!is_mddev_idle(mddev, 0) &&
> + !blk_queue_nonrot(mddev->queue))) {
> msleep(500);
> goto repeat;
> }
> - if (!is_mddev_idle(mddev, 0)) {
> + if (!is_mddev_idle(mddev, 0) &&
> + blk_queue_nonrot(mddev->queue)) {
> /*
> * Give other IO more of a chance.
> * The faster the devices, the less we wait.
> --
> 2.7.5
>
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
