[Openstack-operators] How do you orchestrate external tools (private cloud) ?

gilles.mocellin at nuagelibre.org gilles.mocellin at nuagelibre.org
Tue Jan 12 09:47:21 UTC 2016


I think it will be great to know how operators handle theses kind of 
orchestration :
Add, remove instances and there properties in theses SI tools :
- Monitoring
- Backup

I understand that this use case is certainly specific to private cloud, 
not public ones.

I will be so happy not to rely on an external orchestrator (from an 
OpenStack poit of view).
But I did not find any example where Heat can do these sort of thing.

For the moment, We start to use RunDeck, but I don't like the fact that 
we must use a workflow to create an instance.
If we use OpenStack Directly, we cannot handle constraints we have an 
system names and IPs : two IP addresses by instances, which are 
pre-allocated in our "IPAM" (IPPlan).

The only way I see to be able to use standard OpenStack CLI and API, is 
to handle that orchestration at instance startup.
Which is not easy, mainly if the network configuration must be handled.
There's always possibilities, like :
- a first boot with standard OpenStack DHCP
- a network config handled by cloud-init and/or a configuration 
management system
- reboot in the final good state (but Neutron config like ports will not 
be in sync)

So, how do you orchestrate beside OpenStack ?
Do you abandon OpenStack API compatibility for an orchestrator's one ?
Is there something I haven't seen in Heat ?

More information about the OpenStack-operators mailing list