On 01/05/2018 06:15 AM, Alexey Perevalov wrote: > On 01/03/2018 12:26 AM, Eric Blake wrote: >> On 10/30/2017 08:16 AM, Alexey Perevalov wrote: >>> Postcopy total blocktime is available on destination side only. >>> But query-migrate was possible only for source. This patch >>> adds ability to call query-migrate on destination. >>> To be able to see postcopy blocktime, need to request postcopy-blocktime >>> capability. >> Why not display the stats unconditionally when they are available, >> instead of having to set a capability knob to request them? > That knob necessary to avoid regression if this information > is not necessary, we decided during so long discussion in previous > version of the patch set - it's not necessary always. > But if user requested blocktime and host can't calculate it, > e.g. due to appropriate feature isn't supported in host kernel, > yes, the value will be 0. >
>>> # >>> +# @postcopy-blocktime: total time when all vCPU were blocked during >>> postcopy >>> +# live migration (Since 2.11) >> 2.12 now. >> >> Should this mention the capability knob needed to enable this stat (or >> else get rid of the capability knob and always expose this when >> possible)? >> Okay, so you've explained that the knob is necessary because there is a noticeable performance difference for users that don't care about the statistic; in which case, we DO need a followup patch (or a v2 of Juan's pull request) that documents that this statistic is useless unless you set the migration capability. -- Eric Blake, Principal Software Engineer Red Hat, Inc. +1-919-301-3266 Virtualization: qemu.org | libvirt.org
signature.asc
Description: OpenPGP digital signature