Cirros will be failed to boot correctly with network bridge OVN with certain settings of /etc/hosts.

[TsutomuNakamura]

When I set /etc/hosts (in cloud-init, /etc/cloud/templates/hosts.debian.tmpl) as certain settings, launching Cirros will be failed.

* Message
Exceeded maximum number of retries. Exhausted all hosts available for retrying build failures for instance 0837113a-627d-4935-8a98-a293f10f1804.
* Code
500
* Details
Traceback (most recent call last): File "/usr/lib/python3/dist-packages/nova/conductor/manager.py", line 669, in build_instances raise exception.MaxRetriesExceeded(reason=msg) nova.exception.MaxRetriesExceeded: Exceeded maximum number of retries. Exhausted all hosts available for retrying build failures for instance 0837113a-627d-4935-8a98-a293f10f1804.

Results are difference between settings of /etc/hosts below.

• Settings will be failed

  x.x.x.x controller.example.com controller
  y.y.y.y compute.example.com compute

• Settings will be succeeded

  x.x.x.x controller controller.example.com
  y.y.y.y compute compute.example.com

[TsutomuNakamura]

Differences of /etc/hosts might cause differences amount configurations. Configurations might have differences are.

• controller node

  glance-api.conf
  local_settings.py
  ml2_conf.ini
  neutron.conf
  nova.conf

• compute node

  neutron.conf
  nova.conf

[TsutomuNakamura]

• controller node

• [x] glance-api.conf // both are same

• [x] local_settings.py // both are same

• [x] neutron.conf // both are same

• [x] nova.conf // both are same

• compute node

• [x] neutron.conf // both are same

• [x] nova.conf // both are same

🤔

[TsutomuNakamura]

After deployed with hosts Settings will be failed in first comment of this issue, I changed hosts Settings will be succeeded and re-create Cirros instance, then Cirros failed to running.

A cause of this issue is existing in some roles before a role test_env.

[TsutomuNakamura]

These test cases need modifying /etc/hosts(/etc/cloud/templates/hosts.debian.tmpl) during each border of test cases. A test case need changing /etc/hosts between the role keystone and glance for example.

TestCase01:
  to: keystone, from: glance

Changing border will detect where the bugs are existed.

[TsutomuNakamura]

Test cases

• [x] Case01: keystone - glance
• [x] Case02: glance - placement
• [x] Case03: placement - nova
• [x] Case04: nova - neutron
• [ ] Case05: neutron - horizon ❌

---

Syntax is ... ex) keystone - glance means Run Ansible to a role keystone, then change hosts, then run Ansible to glance.

[TsutomuNakamura]

To detect where the error points are existed, run Ansible till previous point of the Neutron (Nova), then modify sinarios of Neutron, then run Ansible from neutron.

[TsutomuNakamura]

Base source code

• source | diff

Test cases

• ✅ Case11 | Before: source - diff, After: source - diff
• ✅ Case12 | Before - source - diff, After - source - diff
• ✅ Case13 | Before - source - diff, After - source - diff
• ✅ Case14 | Before - source - diff, After - source - diff
• ❌ Case15 | Before - source - diff, After - source - diff

---

There are some problems at installing OVN packages on compute node.

• An instruction of Ansible in this area https://github.com/tsuna-server/build-server-ansible/blob/db3008cbe5992fa97d594cfdfeadda1115a4ed01/roles/neutron/tasks/install_packages/ovn/computes/main.yml#L1-L14

[TsutomuNakamura]

We can set /etc/hosts by writing each hostname at each line to solve this problem temporary.

• Before

  x.x.x.x controller.example.com controller
  y.y.y.y compute.example.com compute

• After

  x.x.x.x controller.example.com
  x.x.x.x controller
  y.y.y.y compute.example.com
  y.y.y.y compute

[TsutomuNakamura]

This already fixed.