why don't you fix it with your configuration management?
You could patch nova but I think is far more easy to tell
puppet/cheff/salt to set live_migration_uri the way you want
my 0.02 $
On 2015-07-02 14:57, Aviolat Romain wrote:
Hi Gustavo,
Yes we already tried to play with the live_migration_uri variable, the problem
is that we use FQDNs as our hypervisors hostnames (it's related to lldp and
networking stuffs).
So as the %s will give me the hypervisor FQDN it will looks like:
live_migration_uri=qemu+tcp://compute0.mydomain.com-api/system
and what I need is:
live_migration_uri=qemu+tcp://compute0-api.mydomain.com/system
AFAIK there is no way to give it the host part of the FQDN only like a %h for
example.
Thanks for your help.
Romain
-----Original Message-----
From: gustavo panizzo (gfa) [mailto:g...@zumbi.com.ar]
Sent: jeudi 2 juillet 2015 08:46
To: Aviolat Romain; openstack@lists.openstack.org
Subject: Re: [Openstack] Openstack live migration and multiple interfaces / fqdn
asuming you are using libvirt as hypervisor you can change this on nova.conf on
the computes
# Migration target URI (any included "%s" is replaced with the # migration
target hostname) (string value) #live_migration_uri=qemu+tcp://%s/system
set something like
live_migration_uri=qemu+tcp://%s-api/system
On 2015-07-02 14:24, Aviolat Romain wrote:
Hi,
I have some troubles with the live migration mechanism of OpenStack and my
setup:
Our compute nodes have two types interfaces with different things running on
top of them.
1GbE interface:
- interface to access the hypervisor OS
- hypervisor access to the core infra services (dns, ntp, syslog, ...). The
FQDN of the compute node is bound to the IP of this interface, let's say
compute0.mydomain.com, compute1.mydomain.com, ...
10 GbE interfaces:
- redundant network, internal to the openstack cloud
- multiple VLANs on top of it to segregate the different services
- api / db
- CEPH cluster network
- CEPH public network
- ...
- A second FQDN is bound to the api / db interface IP address, let's
say compute0-api.mydomain.com
Problem is that when I want to live migrate instance, the Nova driver tries to
connect to the compute0.mydomain.com IP address. It's working but it's far from
ideal (slow network, not redundant, not designed to move internal OpenStack
network flows, ...). Ideally I'd like to be able to use the FQDN that is bound
to the api / db network.
It seems that there is no option on nova to set a specific fqdn for the ssh
live migration, maybe the solution would be to code it ?
Suggestions are welcome !
Thanks for your help,
Romain
Romain Aviolat
Senior System Administrator
R&D and Ops Infrastructure
Kudelski Security - Kudelski Group
Rte de Genève 22-24,
1033 Cheseaux
SWITZERLAND
+41 21 732 03 79
_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe :
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
--
1AE0 322E B8F7 4717 BDEA BF1D 44BB 1BA7 9F6C 6333
keybase: http://keybase.io/gfa
_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
--
1AE0 322E B8F7 4717 BDEA BF1D 44BB 1BA7 9F6C 6333
keybase: http://keybase.io/gfa
_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
