[Openstack-operators] [Nova] instance snapshot failure on mitaka

Jonathan D. Proulx jon at csail.mit.edu
Mon Sep 12 18:52:25 UTC 2016

Hi All,

In my continuing sporadic testing of Kilo->Mitaka upgrade I've run
into an issue with Snapshotting instances.

I'm seeing this Failure on the hypervisor:

2016-09-12 14:14:26.303 24723 ERROR oslo_messaging.rpc.dispatcher libvirtError: internal 
error: unable to execute QEMU command 'migrate': Migration disabled: failed to allocate s
hared memory

This happens wether the instance is running paused or stopped.

I've also tried setting
workrkarounds/disable_libvirt_livesnapshot=False based on

instance ephemeral disk and glance store are RBD backed and using same
credentials.  I am able to upload new glance iamges and nova is able
to 'shelve' the instances which (I think) invokes a similar rbd
snapshot process:

2016-09-12 14:18:45.061 24723 DEBUG nova.virt.libvirt.storage.rbd_utils [] creating snapshot(cbd2266da21d41358259a2e664584299) on rbd image(5b32bb8b-13bf-4b11-882d-3e6183cd1010_disk) create_snap /usr/lib/python2.7/dist-packages/nova/virt/libvirt/storage/rbd_utils.py:380
2016-09-12 14:18:46.344 24723 DEBUG nova.virt.libvirt.storage.rbd_utils [] cloning vms/5b32bb8b-13bf-4b11-882d-3e6183cd1010_disk at cbd2266da21d41358259a2e664584299 to images/2dbedf42-81af-4bf3-b0c4-0a1ba30e3d6a clone /usr/lib/python2.7/dist-packages/nova/virt/libvirt/storage/rbd_utils.py:218
2016-09-12 14:18:46.531 24723 DEBUG nova.virt.libvirt.storage.rbd_utils [] flattening images/2dbedf42-81af-4bf3-b0c4-0a1ba30e3d6a flatten /usr/lib/python2.7/dist-packages/nova/virt/libvirt/storage/rbd_utils.py:267

so looks like it can do the RBD stuff it needs to just fine.

Ubuntu           14.04
libvirt-bin      1.3.1
kernel           3.16.0

lists a similar error but actually related to migration rather than
snapshoting and suggests a newer kernel is the fix, I'm hoping this is
n't required as it woudl really complicate my life to have to
live-migrate-shuffle everything to upgrade kernels prior to upgrading

The answer suggesting this fix is unverified and gives no reason why
this is a fix, so hoping someone has a better answer (better for me



More information about the OpenStack-operators mailing list