[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[blazar] Question about Interaction between Ironic and Blazar


The current architecture of Blazar relies on Nova host aggregates for
physical host reservation.
Unfortunately, Nova host aggregates don't play well with Ironic, as
they are associated with nova-compute services instead of individual
compute hosts.
The Blazar community is well aware of this issue:
We are planning to move to placement aggregates which will remove this
limitation. However, this may only happen in the Train release.

One of the main Blazar deployment is Chameleon
(, which is a bare-metal cloud using
Ironic. Chameleon has used two different workarounds to this issue.
Originally, Nova was modified so that a nova-compute service would
only manage one Ironic bare-metal node. By running as many
nova-compute services as there were bare-metal nodes, host aggregates
could be used by Blazar.
More recently, Chameleon has been using Nova patches developed by Jay
Pipes which allow compute nodes (and thus Ironic nodes) to be
associated with host aggregates (see However, this patch series
was not accepted following discussion at the Rocky PTG in Dublin, with
placement aggregates being preferred.

I am happy to help you adapt Blazar to your requirements. We have a
weekly meeting on Tuesdays, however it's at 0900 UTC which is not
convenient for US timezones. Feel free to join #openstack-blazar on
freenode, I am often online during UTC daytime.

Best wishes,
Pierre Riteau (Blazar PTL for the Stein cycle)

On Wed, 13 Feb 2019 at 18:59, Tzu-Mainn Chen <tzumainn at> wrote:
> Hi! I'm working with both Ironic and Blazar, and came across a strange interaction that I was wondering if the Blazar devs were aware of.
> I had four Ironic nodes registered, and only node A had an instance running on it. I tried adding node B - which was available - to the freepool and got this error:
> 2019-02-13 09:42:28.560 220255 ERROR oslo_messaging.rpc.server
> ERROR: Servers [[{u'uuid': u'298e83a4-7d5e-4aae-b89a-9dc74b4278af', u'name': u'instance-00000011'}]] found for host a00696d5-32ba-475e-9528-59bf11cffea6
> This was strange, because the instance in question was running on node A, and not node B.
> After some investigation, the cause was identified as the following:
> But in the meantime, my question is: have other people using Blazar and Ironic run into this issue? It would seem to imply that Ironic nodes can only be added to the freepool if no instances are created, which poses a long-term maintenance issue. Is there a workaround?
> Thanks,
> Tzu-Mainn Chen