Hi Olivier, this is not really a bug as it's intended. The libvirt_domain_change_vcpus() is not the same as (yet unimplemented) virDomainSetVcpus() API so it's not bug. I realize that naming could be a little confusing so if you like, feel free to rename libvirt_domain_change_vcpus() to libvirt_domain_xml_change_vcpus() as this API should be used to change number of vCPUS in the domain XML itself and implement libvirt_domain_set_vcpus() calling real virDomainSetVcpus() API function.

Thanks, Michal On 09/09/2013 06:03 PM, Olivier Doucet wrote: > Hello everyone, > > I think there is a bug in the current implementation of function > libvirt_domain_change_vcpu(). > When using this function (with second parameter = 2), I have the > following error : > > Error setting maxvcpus=2: (domain_definition):8: Specification mandate > value for attribute curre > <vcpu>2</vcpu> > --^ > > > If I'm right, the equivalent with 'virsh' binary is 'setvcpus', and > this is working. > > If I look into source code > (http://libvirt.org/git/?p=libvirt-php.git;a=blob;f=src/libvirt-php.c;h=7c...) > it seems this operation is done by rewriting the whole XML. > > I think we should use instead C function virDomainSetVcpus > http://libvirt.org/html/libvirt-libvirt.html#virDomainSetVcpus > > > What do you think ? > > Olivier > > -- > libvir-list mailing list > libvir-list(a)redhat.com > https://www.redhat.com/mailman/listinfo/libvir-list

On Mon, Sep 09, 2013 at 06:51:54PM +0200, Michal Novotny wrote: > Hi Olivier, > this is not really a bug as it's intended. The > libvirt_domain_change_vcpus() is not the same as (yet unimplemented) > virDomainSetVcpus() API so it's not bug. I realize that naming could be > a little confusing so if you like, feel free to rename > libvirt_domain_change_vcpus() to libvirt_domain_xml_change_vcpus() as > this API should be used to change number of vCPUS in the domain XML > itself and implement libvirt_domain_set_vcpus() calling real > virDomainSetVcpus() API function. That function is also buggy in that it will silently turn a transient guest into a persistent guest, which is almost certainly not what a caller would expect Daniel