[libvirt] RFC: Exposing "ready" bool (of `query-block-jobs`) or QMP BLOCK_JOB_READY event

Eric Blake eblake at redhat.com
Fri Oct 7 19:01:31 UTC 2016


On 10/07/2016 02:09 AM, Peter Krempa wrote:

>> The existing virDomainGetBlockJobInfo() can't be extended, but it can be
>> fixed to quit reporting cur==end when ready:false.
> 
> Yes, I agree about this one (although I don't really like it [1]), but
> this one will actually fix software not listening for events without any
> change.
> 
> Any new API would not help since the apps would need to change anyways
> thus can use the current correct approach right away even with older
> libvirt versions.
> 
> Peter
> 
> [1]: I'm expecting users to start complaining: "Why is this last byte of
> my image taking so long to copy after the rest copied pretty quickly".

And our response is "We never promised that cur and end are bytes, only
rough status indicators. And we don't know why qemu is taking so long -
move the bug report to them" - if the user can even see this state long
enough for it to bother them.  (Nova is hitting it, because it is a
software-triggered reaction time, not a human in the mix; my
understanding is that it is still at most a fraction of a second where
we'd even have to do the fudging).

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 604 bytes
Desc: OpenPGP digital signature
URL: <http://listman.redhat.com/archives/libvir-list/attachments/20161007/ea83de21/attachment-0001.sig>


More information about the libvir-list mailing list