[libvirt-users] VM migration upon shutdown in centos 7

[Raman Gupta]

Hi,

> I don't see your service being called.
It started when system started, see logs:
Jan 12 14:59:42 desktop4 systemd: systemd 219 running in system mode. (+PAM
+AUDIT +SELINUX +IMA -APPARMOR +SMACK +SYSVINIT +UTMP +LIBCRYPTSETUP
+GCRYPT +GNUTLS +ACL +XZ -LZ4 -SECCOMP +BLKID +ELFUTILS +KMOD +IDN)
Jan 12 14:59:42 desktop4 systemd: Detected architecture x86-64.
Jan 12 14:59:42 desktop4 systemd: Set hostname to <desktop4>.
Jan 12 14:59:42 desktop4 systemd: Started Create list of required static
device nodes for the current kernel.
Jan 12 14:59:42 desktop4 systemd: Started *vPreShutdownHook*.
Jan 12 14:59:42 desktop4 systemd: Starting Create Static Device Nodes in
/dev...

> Anyway, look at libvirt-guests.service file. Looks like the following
lines cause systemd to wait for a command to finish:
The libvirt-guests.service is NOT running in either of my nodes, even
though VM is working fine.

I tried starting  libvirt-guests.service and added it as Requires target in
vPreShutdownHook.service,  but it did not help. On reboot the VM did not
migrated.

Thanks

On Thu, Jan 11, 2018 at 12:56 PM, Michal Privoznik <mprivozn at redhat.com>
wrote:

> [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@/libvirt-guests.sh start
> ExecStop=@libexecdir@/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 at redhat.com>
> > 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
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/libvirt-users/attachments/20180112/b7b1281b/attachment.htm>