controller_node_address needs documentation
Bug #1090991 reported by
Ian Wells
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| Cisco Openstack |
New
|
Undecided
|
Unassigned | ||
| Folsom |
Triaged
|
Medium
|
Unassigned | ||
Bug Description
controller_
- the external API address, used by external processes to get to Horizon and/or the API servers
- the internal communications address, used by compute nodes to communicate with e.g. Rabbit, MySQL
It's not actually documented as either and may in fact be both. What do we want it to be? Decide, fix (if required), document.
Revision history for this message
| Daneyon Hansen (danehans) wrote | #1 |
To post a comment you must log in.
Our current networking model does not give us options for addressing this problem. We need to support multiple interfaces on nodes:
1. API Endpoints (not Keystone Admin though) on Controllers that connect to an API Network
2. Tenant Network interface on all OS nodes used to pass tenant traffic through GRE tunnels (therefore used as GRE endpoints).
3. OOB management network used for CIMC access to servers.
4. OS Management Network used for things like Puppet, MySQL, Keystone Admin API, monitoring tools, etc..