Chapter 4. Compute Nodes in OpenStack Operations Guide

[Lucio Agostinho]

I'm trying live-migration with qemu in a virtualized environment with VirtualBox VMs (not with KVM)

I get mount the NFS among controller and compute nodes, and I can see the instances directory in all compute nodes (this is directory is shared).

I can perform 'migrate' command in terminal and in the dashboard with success. So, the installation seems to be working correctly.

However, the 'live-migration' command is not working in the line command, and no errors are put in the terminal. I'm trying as follow:

# nova live-migration MY_UUID_VM_INSTANCE my_host_destine

No errors are sent to the terminal, and in the dashboard no action is performed after this command.

What I'm doing wrong? I already perform a live migration without openstack using only KVM in physical hosts. I guess that the problem could be the usage of qemu. Someone could help me?

Thanks in advance.

-----------------------------------
Built: 2014-02-14T07:30:04 00:00
git SHA: 833c1e5aa36e234e39aabd7e32f0932d6f292c76
URL: http://docs.openstack.org/trunk/openstack-ops/content/compute_nodes.html
source File: file:/home/jenkins/workspace/openstack-operations-guide/doc/openstack-ops/ch_arch_compute_nodes.xml
xml:id: compute_nodes

[Lucio Agostinho (outrosdiasvirao-n)]

According to openstack documentation: http://docs.openstack.org/trunk/openstack-ops/content/compute_nodes.html

---
Issues with Live Migration

We consider live migration an integral part of the operations of the cloud. This feature provides the ability to seamlessly move instances from one physical host to another, a necessity for performing upgrades that require reboots of the compute hosts, but only works well with shared storage.

Live migration can be also done with non-shared storage, using a feature known as KVM live block migration. While an earlier implementation of block-based migration in KVM and QEMU was considered unreliable, there is a newer, more reliable implementation of block-based live migration as of QEMU 1.4 and libvirt 1.0.2 that is also compatible with OpenStack. However, none of the authors of this guide have first-hand experience using live block migration.

---

So, I'm first-hand experience using live block migration with qemu, and its not working. Someone could help? Thanks.

[Andreas Jaeger (jaegerandi)]

Please ask on ask.openstack.org these question. If you found a solution, documenting it would be welcome ;)

[Lucio Agostinho (outrosdiasvirao-n)]

I solved has follow. This is really a bug in OpenStack Havana:

For a reason that I really do not know, the nova API for libvirt has a option to choose a cache for migrate VM with libvirt. This feature implies in errors using live-migration with --block-migrate that, as default, use cache="none". Changing directly in the source code:

 /usr/lib/python2.6/site-packages/nova/virt/libvirt/driver.py

In the following lines in driver.py:

Line 364:
before: self.disk_cachemodes[disk_type] = cache_mode
after: self.disk_cachemodes[disk_type] = "none"

Near at Line 382:
before: self._disk_cachemode = "writethrough"
after: self._disk_cachemode = "none"

The problem is solved! For a reason that I do not known, I received many bad comments about this problem, what makes difficult to solve it. The manual is poorly documented about live-migration, that really not works with qemu+libvirt without these changes.

[Tom Fifield (fifieldt)]

This appears to be a bug in the code, rather than documentation. I would send it to nova, but unfortunately havana is no longer supported there, so this bug has to be marked as 'wontfix' :|