[pve-devel] [PATCH docs] qm: IO thread: be more precise about how QEMU handles IO

Roland devzero at web.de
Thu Dec 15 11:52:38 CET 2022


>+storage controller, rather than handling all I/O in the main event loop or vCPU
>+threads. This can increase performance, because of improved work distribution.

not only that, it evidently avoids vm locking / cpu stuck/freeze
conditions and vm timing hiccups

see https://bugzilla.kernel.org/show_bug.cgi?id=199727

Am 15.12.22 um 10:59 schrieb Fiona Ebner:
> Reported in the community forum[0].
>
> The setting can already help with a single disk. Without the option,
> there is not one IO thread as the old wording suggested, but IO is
> handled in the main event loop or vCPU threads (see the kvm man page).
>
> [0]: https://forum.proxmox.com/threads/118390/post-518532
>
> Signed-off-by: Fiona Ebner <f.ebner at proxmox.com>
> ---
>   qm.adoc | 11 +++++------
>   1 file changed, 5 insertions(+), 6 deletions(-)
>
> diff --git a/qm.adoc b/qm.adoc
> index f923ccf..fb11e29 100644
> --- a/qm.adoc
> +++ b/qm.adoc
> @@ -249,12 +249,11 @@ Note that *SSD emulation* is not supported on *VirtIO Block* drives.
>   [[qm_hard_disk_iothread]]
>   IO Thread
>   ^^^^^^^^^
> -The option *IO Thread* can only be used when using a disk with the
> -*VirtIO* controller, or with the *SCSI* controller, when the emulated controller
> - type is  *VirtIO SCSI single*.
> -With this enabled, Qemu creates one I/O thread per storage controller,
> -rather than a single thread for all I/O. This can increase performance when
> -multiple disks are used and each disk has its own storage controller.
> +The option *IO Thread* can only be used when using a disk with the *VirtIO*
> +controller, or with the *SCSI* controller, when the emulated controller type is
> +*VirtIO SCSI single*. With *IO Thread* enabled, Qemu creates one I/O thread per
> +storage controller, rather than handling all I/O in the main event loop or vCPU
> +threads. This can increase performance, because of improved work distribution.
>
>
>   [[qm_cpu]]




More information about the pve-devel mailing list