I don't see your service being called.

Anyway, look at libvirt-guests.service file. Looks like the following

[Please keed the list CCed] On 01/10/2018 04:43 PM, Raman Gupta wrote: >> Does this command alone succeed? > Yes. I have used this command to migrate VMs successfully, without even > knowing that spelling has changed. > > >> I don't know enough about systemd but maybe it's not waiting for virsh to > finish? > Yes I also think virsh or libvirtd or related service does not wait > before /root/vm_migrate.sh > is called and hence live migration fails. > If I replace Live Migration with a simple ping to peer node in the > vPreShutdownHook.service, then ping goes thru successfully thus indicating > network service was UP at that time. > >> Can you try to get any logs to see what is going on actually? > Jan 8 15:30:58 desktop4 systemd: Stopping Session c1 of user gdm. > Jan 8 15:30:58 desktop4 gdm: Freeing conversation 'gdm-launch-environment' > with active job > Jan 8 15:30:58 desktop4 systemd: Stopped target Sound Card. > Jan 8 15:30:58 desktop4 systemd: Stopping Sound Card. > Jan 8 15:30:58 desktop4 systemd: Stopping LVM2 PV scan on device 8:3... > Jan 8 15:30:58 desktop4 systemd: Removed slice system-getty.slice. > Jan 8 15:30:58 desktop4 systemd: Stopping system-getty.slice. > Jan 8 15:30:58 desktop4 systemd: Stopping Authorization Manager... > Jan 8 15:30:58 desktop4 systemd: Stopping Virtual Machine and Container > Registration Service... > Jan 8 15:30:58 desktop4 systemd: Stopping GlusterFS brick processes > (stopping only)... > Jan 8 15:30:58 desktop4 systemd: Removed slice > system-selinux\x2dpolicy\x2dmigrate\x2dlocal\x2dchanges.slice. > Jan 8 15:30:58 desktop4 systemd: Stopping > system-selinux\x2dpolicy\x2dmigrate\x2dlocal\x2dchanges.slice. > Jan 8 15:30:58 desktop4 systemd: Stopping Availability of block devices... > Jan 8 15:30:58 desktop4 systemd: Stopping Virtual Machine > qemu-1-GsmController. > Jan 8 15:30:58 desktop4 alsactl[837]: alsactl daemon stopped > Jan 8 15:30:58 desktop4 systemd: Stopping Session 1 of user root. > Jan 8 15:30:58 desktop4 systemd: Stopping Manage Sound Card State (restore > and store)... I don't see your service being called. Anyway, look at libvirt-guests.service file. Looks like the following lines cause systemd to wait for a command to finish: [Service] EnvironmentFile=-/etc/sysconfig/libvirt-guests # Hack just call traditional service until we factor # out the code ExecStart=@libexecdir(a)/libvirt-guests.sh start ExecStop=@libexecdir(a)/libvirt-guests.sh stop Type=oneshot RemainAfterExit=yes StandardOutput=journal+console TimeoutStopSec=0 > > > > On Wed, Jan 10, 2018 at 8:46 PM, Michal Privoznik <mprivozn(a)redhat.com&gt; > wrote: > >> On 01/05/2018 12:00 PM, Raman Gupta wrote: >>> Hi, >>> >>> I have CentOS 7, two node system which allows live VM migration between >>> them. Live migration triggered from virsh is happily happening. I am >> using >>> GlusterFS for replicating VM disk files. >>> >>> Now I want to automatically do the live migration at the time of >>> reboot/shutdown/halt of the host node and for this I have written a >> systemd >>> service unit [vPreShutdownHook.service] and placed the live migration >>> command in a migrate script which is invoked from this service unit. The >>> migrate script is invoked but migration does not happen. >>> If someone has some idea please help me to migrate VM upon shutdown. >>> >>> >>> ######## vPreShutdownHook.service ############### >>> >>> [Unit] >>> Description=vPreShutdownHook >>> Requires=network.target >>> Requires=libvirtd.service >>> Requires=dbus.service >>> Requires=glusterd.service >>> Requires=glusterfsd.service >>> DefaultDependencies=no >>> Before=shutdown.target reboot.target >>> >>> [Service] >>> Type=oneshot >>> RemainAfterExit=true >>> ExecStart=/bin/true >>> ExecStop=/root/vm_migrate.sh >>> >>> [Install] >>> WantedBy=multi-user.target >>> >>> >>> ########## Command to migrate ############ >>> /usr/bin/virsh migrate --verbose --p2p --tunneled --live --compressed >>> --comp-methods "mt" --comp-mt-level 5 --comp-mt-threads 5 >>> --comp-mt-dthreads 5 MY_VM qemu+ssh://root@$node2/system >> >> Does this command alone succeed? >> BTW: unless really needed --live will only make the migration take longer. >> >> I don't know enough about systemd but maybe it's not waiting for virsh >> to finish? Can you try to get any logs to see what is going on actually? >> >> Michal >> > Michal