Hide Forgot
=Comment: #0================================================= Emily J. Ratliff <ratliff@austin.ibm.com> - 1. Feature Overview: Feature Id: [202014] a. Name of Feature: Add async qemu machine protocol (qmp) to libvirt b. Feature Description Switch libvirt to use qmp to manipulate qemu/kvm VMs 2. Feature Details: Sponsor: LTC Architectures: x86 x86_64 Arch Specificity: Both Affects Kernel Modules: Yes Delivery Mechanism: Direct from community Category: Xen Request Type: Driver - New - Open Source d. Upstream Acceptance: Submitted Sponsor Priority 2 f. Severity: Medium IBM Confidential: no Code Contribution: 3rd party code g. Component Version Target: latest libvirt 3. Business Case Switching to qmp provides a more reliable protocol on top of which we can build a more robust, asynchronous communication channel for managing VMs. The result provides management applications which consume libvirt greater control over VM bahavior. 4. Primary contact at Red Hat: John Jarvis jjarvis@redhat.com 5. Primary contacts at Partner: Project Management Contact: Stephanie Glass, sglass@us.ibm.com Technical contact(s): Ryan Harper, raharper@us.ibm.com IBM Manager: Warren Grunbok II, grunbok@us.ibm.com
> 3. Business Case > Switching to qmp provides a more reliable protocol on top of which we can build > a more robust, asynchronous communication channel for managing VMs. The result > provides management applications which consume libvirt greater control over VM > behaviour. Since the QEMU machine protocol is a private internal implementation detail of libvirt's QEMU driver, merely changing the protocol has no effect on functionality exposed to applications via the libvirt API. Although QMP looks to be promising as a long term enhancement to QEMU, its upstream development is still ongoing & thus is not scheduled for inclusion in the next official QEMU release. In addition, the libvirt side of support for this new QMP protocol has not commenced development upstream, pending finalization of the QEMU code support for QMP. While the existing monitor protocol has a number of limitations, it is clearly the most tried & tested reliable approach at this point in time. Given the ongoing upstream development work, I do not think that QMP will be sufficiently mature & widely tested enough to be considered suitable for inclusion RHEL in the near immediate future. Can you give further details on exactly what application facing functionality you wish to see in libvirt which is believed to depend on use of this new protocol ? It may well be practical to provide the required enhancements in libvirt's QEMU support, without switching to the QMP.
------- Comment From hollisb@us.ibm.com 2009-08-19 12:21 EDT------- The real requirement is to handle asynchronous watchdog events. If these can be transmitted from qemu to libvirt (and then emitted as domain events) in some way other than QMP, that's good enough.
Posting patches upstream will certainly help getting there if you expect this for 5.5, Daniel
------- Comment From hollisb@us.ibm.com 2009-10-01 17:41 EDT------- Changing summary, since QMP isn't the real requirement (as discussed here).
------- Comment From hollisb@us.ibm.com 2009-10-06 17:19 EDT------- This won't be ready in time for 5.5.
------- Comment From sglass@us.ibm.com 2009-10-07 10:44 EDT------- Deferring this feature from RHEL 5.5 for now.
setting CLOSED DEFERRED to match IBM status.