Description of problem:Inconsistent error msg when nonexistent numa nodeset defined in numatune Version-Release number of selected component (if applicable): # rpm -q libvirt qemu-kvm libvirt-8.0.0-22.module+el8.9.0+19544+b3045133.x86_64 qemu-kvm-6.2.0-37.module+el8.9.0+19491+15e62c0a.x86_64 How reproducible: 100% Steps to Reproduce: 1. Prepare a NUMA host # numactl --hard available: 2 nodes (0-1) node 0 cpus: 0 1 2 node 0 size: 5725 MB node 0 free: 5061 MB node 1 cpus: 3 4 5 node 1 size: 5990 MB node 1 free: 5684 MB node distances: node 0 1 0: 10 20 1: 20 10 2. Start a guest with nonexistent numa nodeset defined in numatune setting: <numatune> <memory mode="strict" nodeset="1-2"/> </numatune> 3. Error msg prompts when start the guest # virsh start vm1 error: Failed to start domain 'vm1' error: Invalid value '1-2' for 'cpuset.mems': Invalid argument 4. If start a guest with nonexistent numa nodeset defined in numatune and numa topology setting <numatune> <memory mode="strict" nodeset="1-2"/> <!-- or with restrictive mode--> </numatune> ... <numa> <cell id='0' cpus='0' memory='1048576' unit='KiB' /> <cell id='1' cpus='1' memory='1048576' unit='KiB' /> </numa> ... 5. Error msg prompts when start the guest # virsh start vm1 error: Failed to start domain 'vm1' error: unsupported configuration: NUMA node 2 is unavailable Actual results: Nonexistent numa nodeset defined in numatune with and without numa topology have different error messages. Expected results: Nonexistent numa nodeset defined in numatune without numa topology should have the same meaningful error message as numa topology defined. Additional info: When Nonexistent numa nodeset defined in numatune without numa topology, if set numatune mode with "preferred" and "interleave" mode the error msg is like below: # virsh start vm1 error: Failed to start domain 'vm1' error: internal error: Process exited prior to exec: libvirt: error : unsupported configuration: NUMA node 2 is unavailable Even though we could see the error, the whole output is different with numa topology defined.
Is this reproducible with RHEL-9?
(In reply to Jaroslav Suchanek from comment #1) > Is this reproducible with RHEL-9? No, rhel9 doesn't have this issue.