On 17/08/18 14:09 -0500, Jay S Bryant wrote:
On 8/17/2018 1:34 PM, Sean McGinnis wrote:
Has there been a discussion on record of how use of placement by cinder
would affect "standalone" cinder (or manila) initiatives where there is a
desire to be able to run cinder by itself (with no-auth) or just with
keystone (where OpenStack style multi-tenancy is desired)?
Tom Barron (tbarron)
A little bit. That would be one of the pieces that needs to be done if we were
to adopt it.
Just high level brainstorming, but I think we would need something like we have
now with using tooz where if it is configured for it, it will use etcd for
distributed locking. And for single node installs it just defaults to file
locks.
Sean and Tom,
That brief discussion was in Vancouver:
https://etherpad.openstack.org/p/YVR-cinder-placement
Thanks, Jay.
But as Sean indicated I think the long story short was that we would
make it so that we could use the placement service if it was available
but would leave the existing functionality in the case it wasn't
there.
I think that even standalone if I'm running a scheduler (i.e., not
doing emberlib version of standalone) then I'm likely to want to run
them active-active on multiple nodes and will need a solution for the
current races. So even standalone we face the question of do we use
placement to solve that issue or do we introduce some coordination
among the schedulers themselves to solve it.
-- Tom Barron (tbarron)
Jay
__________________________________________________________________________
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
__________________________________________________________________________
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