On Tue, May 15, 2018 at 3:47 PM, Ilya Dryomov <idryomov@xxxxxxxxx> wrote:
> On Tue, May 15, 2018 at 11:03 AM, David Disseldorp <ddiss@xxxxxxx> wrote:
>> On Mon, 14 May 2018 06:27:00 -0700, Jason Dillaman wrote:
>>
>>> While I cannot speak for the upstream kernel maintainers, in the past
>>> they have rejected the RBD module for LIO and you might have similar
>>> issues wiring timestamp changes into tcp_recvmsg. Plus, in a large
>>> deployment where you have lots of initiators connecting to lots of
>>> targets, I really wonder what benefit you will receive from
>>> active/active versus just level-loading the targets and not having to
>>> deal w/ the added racy complexities.
>>
>> Thanks for the feedback, Jason. In this case, the PoC is RBD and OSD
>> only; there's no LIO involvement. The idea is that if the changes are
>> considered useful / non-intrusive for standalone kRBD then I'll proceed
>> with something similar for librbd and tcmu-runner.
>
> Are you asking from a PoC point of view or with an eye towards
> upstreaming? krbd changes look reasonable in principle, but I don't
> see how any of it is useful without LIO changes. Doesn't the whole
> thing boil down to where you set the expiration time? AIUI if you do
> it somewhere in rbd, it's too late.
... or rather where you get the expiration time from. Setting it in
rbd_img_request_create() is obviously fine, it is taking the current
time in rbd_queue_workfn() and passing it to rbd_img_request_create()
what I'm getting at.
Thanks,
Ilya
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
