[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [libvirt] [RFC PATCH 00/12] Add block write threshold event



On Thu, Feb 23, 2017 at 13:50:03 -0600, Eric Blake wrote:
> On 02/23/2017 01:21 PM, Peter Krempa wrote:
> > Since I'm a procrastinator and I wanted to do this after full node-name support
> > is added I'm kind of late for this release here. Since oVirt is asking for this
> > feature for a very long time I would like to deliver it ASAP though.
> 
> Thanks for picking up on this. I tried once a long time ago:
> https://www.redhat.com/archives/libvir-list/2015-June/msg01111.html
> 
> and got stalled on node name weirdness, so it never went in.
> 
> > 
> > This series adds a event which is fired once a guest writes beyond a configured
> > offset in the backing file. The threshold offset is configurable via a new API.
> > 
> > This series is in RFC state since I did not manage to polish few details mostly
> > connected to node names:
> > 
> > - currently only the top level image can be selected
> > - some blockjobs may clear the remembered node names and thus break this
> >   ( re-detection of the backing chain needs to be connected with node name
> >     detection )
> > - the node name detection code is VERY crude
> 
> Indeed, that's where I got stalled, but I think you may already be
> further along than my attempt. (It may also be that the QMP side from
> qemu is easier to use now than when I first tried)
> 
> > - I'm not sure how well it plays with qcow images, since I select only the top
> >   level node (qcow protocol) rather than the file backing it
> 
> That may be a deal-breaker - my recollection is that you HAVE to use the
> node-name of the file backing the qcow2 to get the useful threshold, and

Yes, that's what I expect as well.

> not the node-name of the qcow2 protocol.

For -blockdev support we will need full node-name detection anyways.
I plan to keep two node-names per virStorageSource, one that describes
the format layer and one that describes the protocol/backing layer. The
last patch is just a very dirty hack to see that the rest was wired up
properly.

> 
> > - the documentation probably sucks since I did not proof-read
> > 
> > Any feedback is welcome.
> > 
> > Peter Krempa (12):
> >   util: storage: Split out useful bits of virStorageFileParseChainIndex
> >   util: storage: Add preliminary storage for node names into
> >     virStorageSource
> >   lib: Introduce event for tracking disk backing file write threshold
> >   qemu: monitor: Add support for BLOCK_WRITE_THRESHOLD event
> >   qemu: domain: Add helper to lookup disk by node name
> >   qemu: domain: Add helper to generate indexed backing store names
> >   qemu: process: Wire up firing of the
> >     VIR_DOMAIN_EVENT_ID_BLOCK_THRESHOLD event
> >   lib: Add API for setting the threshold size for
> >     VIR_DOMAIN_EVENT_ID_BLOCK_THRESHOLD
> >   virsh: Implement 'blockthreshold' command to call
> >     virDomainSetBlockThreshold
> >   qemu: domain: Add helper to look up disk soruce by the backing store
> >     string
> >   qemu: implement qemuDomainSetBlockThreshold
> >   qemu: WIP: lookup nodenames
> 
> -- 
> Eric Blake   eblake redhat com    +1-919-301-3266
> Libvirt virtualization library http://libvirt.org
> 




> --
> libvir-list mailing list
> libvir-list redhat com
> https://www.redhat.com/mailman/listinfo/libvir-list

Attachment: signature.asc
Description: PGP signature


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]