Red Hat Bugzilla – Bug 1326270
Migration failed when setting vnc_auto_unix_socket = 1
Last modified: 2016-11-03 14:42:01 EDT
Created attachment 1146316 [details] libvirtd log on both the source host and target host, and qemu log on target host Description of problem: 1) Migration failed when setting vnc_auto_unix_socket = 1 and current occupied domain ids on source host and target host are different.(e.g. the max occupied domain id is 4 on source host, while the max occupied domain id is 1 on target host) # virsh migrate rhel7.2-1030 qemu+ssh://10.66.4.113/system --live --verbose error: internal error: process exited while connecting to monitor: 2016-04-12T07:54:18.328793Z qemu-kvm: -vnc unix:/var/lib/libvirt/qemu/domain-4-rhel7.2-1030/vnc.sock: Failed to start VNC server: Failed to bind socket to /var/lib/libvirt/qemu/domain-4-rhel7.2-1030/vnc.sock: No such file or directory 2) Restore a managed saved domain failed. # virsh managedsave rhel7.2-1030 # virsh start rhel7.2-1030 error: Failed to start domain rhel7.2-1030 error: internal error: process exited while connecting to monitor: 2016-04-12T08:35:53.536685Z qemu-kvm: -vnc unix:/var/lib/libvirt/qemu/domain-4-rhel7.2-1030/vnc.sock: Failed to start VNC server: Failed to bind socket to /var/lib/libvirt/qemu/domain-4-rhel7.2-1030/vnc.sock: No such file or directory Version-Release number of selected component: libvirt-1.3.3-1.el7.x86_64 How reproducible: 100% Steps to Reproduce: 0. Set vnc_auto_unix_socket = 1 on both source and target hosts, restart libvirtd service. Start different number of guests on source and target hosts, to make the current occupied domain id different, e.g. start three guests on source host, and start one guest on target host. 1. Set guest graphic type=vnc: # virsh dumpxml rhel7.2-1030 ... <graphics type='vnc' port='-1' autoport='yes'/> ... 2. Start guest rhel7.2-1030 on source host(its domain id will be 4) and migrate it to target host(on target, domain id of guest rhel7.2 will be 2): # virsh migrate rhel7.2-1030 qemu+ssh://10.66.4.113/system --live --verbose error: internal error: process exited while connecting to monitor: 2016-04-12T07:54:18.328793Z qemu-kvm: -vnc unix:/var/lib/libvirt/qemu/domain-4-rhel7.2-1030/vnc.sock: Failed to start VNC server: Failed to bind socket to /var/lib/libvirt/qemu/domain-4-rhel7.2-1030/vnc.sock: No such file or directory 3. Do managedsave and restore: # virsh managedsave rhel7.2-1030 # virsh start rhel7.2-1030 error: Failed to start domain rhel7.2-1030 error: internal error: process exited while connecting to monitor: 2016-04-12T08:35:53.536685Z qemu-kvm: -vnc unix:/var/lib/libvirt/qemu/domain-4-rhel7.2-1030/vnc.sock: Failed to start VNC server: Failed to bind socket to /var/lib/libvirt/qemu/domain-4-rhel7.2-1030/vnc.sock: No such file or directory Actual results: Migration failed. Restore a managed saved domain failed. Expected results: Migration can succeed. And domain can restore from managed saved state successfully.
Fix posted upstream: https://www.redhat.com/archives/libvir-list/2016-April/msg01733.html
Fixed upstream with v1.3.4-rc1-6-g55320c23dd16: commit 55320c23dd163e75eb61ed6bea2f339ccfeff4f9 Author: Martin Kletzander <mkletzan@redhat.com> Date: Tue Apr 26 14:27:16 2016 +0200 qemu: Regenerate VNC socket paths
Reproduce on build libvirt-1.3.3-1.el7.x86_64, verify pass on build libvirt-1.3.4-1.el7.x86_64. Steps: scenario1: 0. Set vnc_auto_unix_socket = 1 on both source and target hosts, restart libvirtd service. Start different number of guests on source and target hosts, to make the current occupied domain id different, e.g. start three guests on source host, and start one guest on target host. 1. Prepare a runnig guest with graphic type=vnc: # virsh dumpxml rhel7.2 ... <graphics type='vnc' socket='/var/lib/libvirt/qemu/domain-3-rhel7.2/vnc.sock'> ... 2.Do migrate: #virsh migrate rhel7.2 qemu+ssh://10.66.4.148/system --live --verbose 3.After migration completed, check the vnc socket paths in the target host: #virsh dumpxml rhel7.2 | grep vnc ... <graphics type='vnc' socket='/var/lib/libvirt/qemu/domain-2-rhel7.2/vnc.sock'> ... scenario2: 1.set vnc_auto_unix_scoket=1 and restart libvirtd service; 2. Prepare a runnig guest with graphic type=vnc: # virsh dumpxml rhel7.2 ... <graphics type='vnc' socket='/var/lib/libvirt/qemu/domain-3-rhel7.2/vnc.sock'> ... 3.Do managedsave: #virsh managedsave rhel7.2 Domain rhel7.2 state saved by libvirt 4.Start the guest: #virsh start rhel7.2 Domain rhel7.2 started 5.Check the vnc socket paths: #virsh dumpxml rhel7.2 ... <graphics type='vnc' socket='/var/lib/libvirt/qemu/domain-4-rhel7.2/vnc.sock'> ...
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, 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://rhn.redhat.com/errata/RHSA-2016-2577.html