Description of problem: libvirt's NUMA configuration may result in a command line which qemu considers deprecated: <cpu> <topology sockets='2' cores='4' threads='2'/> <numa> <cell cpus='0-7' memory='109550' unit='KiB'/> <cell cpus='8-15' memory='109550' unit='KiB'/> </numa> </cpu> -numa node,nodeid=0,cpus=0-7,mem=107 \ -numa node,nodeid=1,cpus=8-15,mem=107 \ qemu then reports: 2019-12-10T09:47:19.571513Z qemu-kvm: -numa node,nodeid=0,cpus=0-7,mem=107: warning: Parameter -numa node,mem is deprecated, use -numa node,memdev instead Version-Release number of selected component (if applicable): libvirt version: 5.6.0, package: 6.module+el8.1.0+4244+9aa4e6bb qemu version: 4.1.0qemu-kvm-4.1.0-14.module+el8.1.0+4548+ed1300f4 How reproducible: Steps to Reproduce: 1. 2. 3. Actual results: Expected results: Additional info:
Design discussion started here: https://www.redhat.com/archives/libvir-list/2020-January/msg00709.html
(In reply to Michal Privoznik from comment #1) > Design discussion started here: > > https://www.redhat.com/archives/libvir-list/2020-January/msg00709.html Was it fixed upstream already?
Unfortunately, no.
Patches posted upstream: https://www.redhat.com/archives/libvir-list/2020-May/msg00490.html
v2: https://www.redhat.com/archives/libvir-list/2020-May/msg00676.html
Merged upstream: aad31f46fb qemuBuildNumaArgStr: Use modern -numa memdev= if old -numa mem= is unsupported 0246a4f0a7 qemuBuildNumaArgStr: Switch order of if() and for() bba05b01ce qemu: Track numa-mem-supported machine attribute v6.3.0-103-gaad31f46fb
Verified with libvirt-6.6.0-2.module+el8.3.0+7567+dc41c0a9.x86_64 & qemu-kvm-5.1.0-2.module+el8.3.0+7652+b30e6901.x86_64 Start a domain with numa part xml: <cpu mode='custom' match='exact' check='none'> <model fallback='forbid'>qemu64</model> <topology sockets='2' dies='1' cores='4' threads='2'/> <numa> <cell id='0' cpus='0-7' memory='109550' unit='KiB'/> <cell id='1' cpus='8-15' memory='109550' unit='KiB'/> </numa> </cpu> # virsh start vm1 Check the qemu command line: -object memory-backend-ram,id=ram-node0,size=112197632 -numa node,nodeid=0,cpus=0-7,memdev=ram-node0 -object memory-backend-ram,id=ram-node1,size=112197632 -numa node,nodeid=1,cpus=8-15,memdev=ram-node1
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory (virt:8.3 bug fix and enhancement update), and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2020:5137
Hi, Seems libvirt is using the old cmd line again. E.g. <cpu mode='host-model' check='partial'> <numa> <cell id='0' cpus='0-3' memory='1048576' unit='KiB'/> </numa> </cpu> The qemu cli generated is '-numa node,nodeid=0,cpus=0-3,mem=1024'. Packages: libvirt-client-7.4.0-1.module+el8.5.0+11218+83343022.x86_64 qemu-kvm-6.0.0-21.module+el8.5.0+11555+e0ab0d09
I can't reproduce it in the same version . Is the libvirt upgraded from some lower version in your environment?
(In reply to Jing Qi from comment #15) > I can't reproduce it in the same version . > Is the libvirt upgraded from some lower version in your environment? No, it's fresh installed. BTW, I noticed the old cmd line is used in early versions as the warning message in https://bugzilla.redhat.com/show_bug.cgi?id=1938956#c0.
The use of old/new cmd line is dependent on machine type. I mean, when libvirt queries QEMU capabilities, it looks for 'numa-mem-supported' attribute, e.g. like this: {"hotpluggable-cpus": true, "name": "pc-q35-5.2", "numa-mem-supported": false, "default-cpu-type": "qemu64-x86_64-cpu", "cpu-max": 288, "deprecated": false, "default-ram-id": "pc.ram"} and only if numa-mem-supported is true then the old -numa mem=X is used (IIRC there were some problems with guest ABI, hence machine type dependency). I think pc-q35-5.1 and pc-i440fx-5.1 were the first machine types that started reporting numa-mem-supported:false.
(In reply to Michal Privoznik from comment #17) > The use of old/new cmd line is dependent on machine type. I mean, when > libvirt queries QEMU capabilities, it looks for 'numa-mem-supported' > attribute, e.g. like this: > > {"hotpluggable-cpus": true, "name": "pc-q35-5.2", "numa-mem-supported": > false, "default-cpu-type": "qemu64-x86_64-cpu", "cpu-max": 288, > "deprecated": false, "default-ram-id": "pc.ram"} > > and only if numa-mem-supported is true then the old -numa mem=X is used > (IIRC there were some problems with guest ABI, hence machine type > dependency). I think pc-q35-5.1 and pc-i440fx-5.1 were the first machine > types that started reporting numa-mem-supported:false. Yes, it's using the new cmd line with q35 machine types. Thanks Michal for the explanation!