If the build host has two interfaces things won't work
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Cisco Openstack |
Fix Committed
|
Undecided
|
Unassigned | ||
Folsom |
Fix Released
|
High
|
Ian Wells |
Bug Description
Typically I configure a build node with two interfaces and routing disabled. One interface faces the internet; one is the interface to which the machines to install are attached.
In this case, the build node has two IPs, two names and two domains:
build-node.external = a.b.c.d - the external interface
build-node.internal = e.f.g.h - the PXE-boot interface
In this case, you would need to set
$build_node_fqdn = 'build-
$domain_name = 'internal' (perhaps not totally obvious)
Your puppet certificate must also be signed with the *internal* address of the cobbler node, since that's the one that the installation nodes will see. I have had many fights trying to get that to happen in the past.
I propose that we do not insist that the user provide an FQDN for the build_node:
$build_node = 'build-server'
and we construct the correct FQDN where required so that we can use the appropriate one as needed. I also recommend that we use the internal one in all cases where the communication is from or with an installation host, which will work correctly for both single-interface and multi-interface installations.
Changed in openstack-cisco: | |
milestone: | none → 2012.2.2 |
Changed in openstack-cisco: | |
milestone: | 2012.2.2 → none |
This looks suspiciously like Bug 1090336 (or at least like the proposed fixes are similar). Should these be duped?