On 1/26/2017 8:40 AM, Matt Riedemann wrote:
And circling back on *that*, we've agreed to introduce a new service
version for the compute to indicate it's Ocata or not. Then we'll:
* check in the scheduler if the minimum compute service version is ocata,
* if minimum is ocata, then use placement, else fallback to the old
resource tracker data in the compute_nodes table - then we remove that
fallback in Pike.
We'll also have a check for the placement config during init_host on the
ocata compute such that if you are upgrading to ocata code for the
compute but don't have placement configured, it's a hard fail and the
nova-compute service is doing to die.
I'm pretty sure we've come full circle on this now.
Just an update:
The two nova changes for the filter scheduler using placement are
passing CI testing now and are approved. They are held up from being
merged due to the grenade change at the bottom of the series which
installs placement before upgrading to ocata and configures nova-compute:
https://review.openstack.org/424730
We need that in as soon as possible, so Sean, if you're reading this,
you know what to do.
--
Thanks,
Matt Riedemann
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
