Create VMs with proper FQDN instead of shortname only #2151
Conversation
|
Skipping CI for Draft Pull Request. |
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here.
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
Build failed (check pipeline). Post https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/5aaff74c058842819528087fc9206c82 ✔️ openstack-k8s-operators-content-provider SUCCESS in 1h 44m 03s |
cjeanner
force-pushed
the
libvirt_manager/use-fqdn
branch
from
f9fce14 to
27771cf
Compare
|
Build failed (check pipeline). Post https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/566df85bfdcf4b08a0c852d6ac5b7008 ✔️ openstack-k8s-operators-content-provider SUCCESS in 2h 39m 04s |
cjeanner
force-pushed
the
libvirt_manager/use-fqdn
branch
from
27771cf to
eded598
Compare
cjeanner
changed the title
cjeanner
force-pushed
the
libvirt_manager/use-fqdn
branch
from
eded598 to
c70c24f
Compare
cjeanner
force-pushed
the
libvirt_manager/use-fqdn
branch
from
c70c24f to
6dfb897
Compare
playbooks/06-deploy-architecture.yml
Outdated
| ] | ||
| ) if (_ipv4_network_data | length > 0) else [] | ||
| }} | ||
| rescue: |
There was a problem hiding this comment.
| ansible.builtin.shell: | ||
| executable: /bin/bash | ||
| cmd: |- | ||
| test -d /home/zuul && exit 0; | ||
| set -xe -o pipefail; | ||
| echo "{{ vm }}" | sudo tee /etc/hostname; | ||
| sudo hostname -F /etc/hostname; | ||
| sudo hostnamectl hostname "{{ _fqdn }}" |
There was a problem hiding this comment.
Yes, let hostnamectl set it.
There was a problem hiding this comment.
the canonical hostname of the edpm node if its hosting a nova is required to be the ctrlplane fqdn
so cifmw_reproducer_domain must always equal the ctlplane network domain in the network atachmetn definition.
without this it will break live migration and tls and it will cause issues with adoption potentially too.
There was a problem hiding this comment.
If you set hostname with fqdn, and use the same in nodeset spec node hostName, it would be used as canonical_hostname and resolve to the ctlplane network ip. I guess you're not updating the nodeset spec with the correct hostName.
|
This looks good to me. Let's get it green. |
|
/LGTM |
|
Let's make it passes all downstream tests before applying the Looks Good to Me |
cjeanner
force-pushed
the
libvirt_manager/use-fqdn
branch
5 times, most recently
from
3069ac6 to
a4ee079
Compare
|
@fultonj Status update on the testing side:
The expected ceph spec is now showing FQDN such as: [root@compute-0 specs]# grep -B1 -A3 hosts ceph_spec.yaml
placement:
hosts:
- compute-0.ocp.openstack.lab
- compute-1.ocp.openstack.lab
- compute-2.ocp.openstack.lab
--
placement:
hosts:
- compute-0.ocp.openstack.lab
- compute-1.ocp.openstack.lab
- compute-2.ocp.openstack.lab
--
placement:
hosts:
- compute-0.ocp.openstack.lab
- compute-1.ocp.openstack.lab
- compute-2.ocp.openstack.labAnd it seems to match the applied spec: [root@compute-0 specs]# ceph orch ls --export | grep hosts -A 3
hosts:
- compute-0.ocp.openstack.lab
- compute-1.ocp.openstack.lab
- compute-2.ocp.openstack.lab
--
hosts:
- compute-0.ocp.openstack.lab
- compute-1.ocp.openstack.lab
- compute-2.ocp.openstack.lab
--
hosts:
- compute-0.ocp.openstack.lab
- compute-1.ocp.openstack.lab
- compute-2.ocp.openstack.lab
--
hosts:
- compute-0.ocp.openstack.lab
- compute-1.ocp.openstack.lab
- compute-2.ocp.openstack.lab
--
hosts:
- compute-0.ocp.openstack.lab
- compute-1.ocp.openstack.lab
- compute-2.ocp.openstack.labSome more status data, to ensure services are all good: Note: this is from my va-hci deployment, I'll do the same checks on the delta-v6. But since it's hitting tempest, I'd expect it to be all happy as well. |
|
Looks promising. Let's plan to get this merged early next week. Thank you @cjeanner |
cjeanner
force-pushed
the
libvirt_manager/use-fqdn
branch
from
a4ee079 to
85bca7c
Compare
There is a need to get unique hostnames for the VM that will be mounting some shared volumes in some infrastructure. Allowing to get FQDN instead of shortnames allows to change the domain of the VMs, making FQDN actually unique for each run. The way we will leverage that feature is still unknown, but the support will be in place whenever we need it. A possibility would be to set the cifmw_devscripts_config.base_domain to a unique value, or directly set the cifmw_reproducer_domain to whatever value the user wants.
cjeanner
force-pushed
the
libvirt_manager/use-fqdn
branch
from
85bca7c to
5e1abd0
Compare
|
Notes:
Otherwise, it seems to be a plain success. There might raise some issues for some cases where tests/others were using some inconsistent patterns, "hoping" it would work (that was the case for neutron whitebox), but we should have squashed most of the issues beforehand. |
|
The patches under https://review.opendev.org/q/topic:%22use-startswith%22 have merged. I assume the plan to run tests again to confirm that the merges produced the expected results and then if it's green to merge this patch. |
There may be other issues on the nova side with live migration, I was told. |
|
Leading to far too many issues with FQDN inconsistency in multi-NIC environment. With the DCN and spine/leaf case, we can't reliably generate things properly. Dropping this PR, already lost far too much time on it with little to no benefit overall. To get unique hostnames, I'll modify the libvirt_manager to introduce some random string in the shortname, and correct the various matches/extractions to take that into account. |
There is a need to get unique hostnames for the VM that will be mounting
some shared volumes in some infrastructure.
Allowing to get FQDN instead of shortnames allows to change the domain
of the VMs, making FQDN actually unique for each run.
The way we will leverage that feature is still unknown, but the support
will be in place whenever we need it.
A possibility would be to set the cifmw_devscripts_config.base_domain to
a unique value, or directly set the cifmw_reproducer_domain to whatever
value the user wants.
Testing (not in commit message)