Hide Forgot
Description of problem: start the virtual machine vm 1 with macvtap net device, on another host there is one same virtual machine vm2, and vm2 can ping packets to vm1. When vm1 live migrates to the host which has the virtual machine vm2, the time of ping interuption is too long. Version-Release number of selected component (if applicable): 1.2.21 How reproducible: Steps to Reproduce: 1. Start the virtual machine vm 1 with macvtap net device 2. On another host, start one same virtual machine vm2 3. Vm2 can ping packets to vm1 4. Vm1 does live migration to the host which has the virtual machine vm2 Actual results: the time of ping interuption is too long Expected results: Additional info:
What distro are you using? What guest OS? What does 'too long' mean specifically? How long does the migration itself take?
(In reply to Cole Robinson from comment #1) > What distro are you using? What guest OS? > What does 'too long' mean specifically? How long does the migration itself > take? Thanks for your reply. The host OS is centos7.0, the same to the guest. Libvirt version is 1.2.21. The time of ping interuption is too long, this 'too long' takes about 200 seconds. About the migration time, I will check again and reply to you.
Where does libvirt 1.2.21 come from? The official centos versions are different AFAICT. If you can try a newer libvirt version that would help (this bug tracker tends to only track latest libvirt versions)
As far as I can tell, CentOS 7.0 has libvirt 1.1.1. This sounds eerily similar to a bug that was fixed in libvirt 1.2.12 - Bug 1081461. Those patches were also backported to the RHEL libvirt version 1.2.8-11 which was in RHEL/CentOS 7.1. Is there a chance you're misreading your version number?
EOL