A related question is, does the Keystone service name have any significance other than as a label. For example, could/would you create the following services in Keystone for a Cloud Controller with 2 Compute nodes?
Service Name Service Type nova_cc nova nova_c1 nova nova_c2 nova glance image keystone identity
Then would you associate a single endpoint, i.e. each host, to each service?
nova_cc endpoint at 10.10.10.11 nova_c1 endpoint at 10.10.10.12 nova_c2 endpoint at 10.10.10.13 glance endpoint at 10.10.10.14 keystone endpoint at 10.10.10.11
Or would you associate multiple endpoints to one service?
nova nova glance image keystone identity
nova endpoint at 10.10.10.11 nova endpoint at 10.10.10.12 nova endpoint at 10.10.10.13 glance endpoint at 10.10.10.14 keystone endpoint at 10.10.10.11
Hi Bill,
Generally speaking, you'd only create a endpoints for any separate instances running the nova-api services. Many deployments run multiple of these (for fault tolerance and redundancy) behind a load balancer (such as an F5 or HAproxy) and define only a single endpoint at the load balancer's relevant IP address/hostname.
The current run of OpenStack clients are not very intelligent, and generally only look for the first publicURL endpoint they can find to interact with the service at hand, so I'd generally recommend this setup with the Essex release.
Asked: 2012-06-18 15:56:47 -0500
Seen: 550 times
Last updated: Jun 24 '12
