when managing /etc/hosts, remove existing host entries

[Radosław Piliszek]

When kolla-ansible is to manage /etc/hosts, it adds its entries but it does not ensure current ones are gone which leads to RabbitMQ issues.

Example /etc/hosts contents after run:

127.0.0.1 localhost
::1 localhost localhost.localdomain localhost6 localhost6.localdomain6
172.16.34.10 cloudnode1
172.16.34.11 cloudnode2
172.16.34.12 cloudnode3
172.16.34.240 kolladeploy
# BEGIN ANSIBLE GENERATED HOSTS
172.16.35.10 cloudnode1
172.16.35.11 cloudnode2
172.16.35.12 cloudnode3
# END ANSIBLE GENERATED HOSTS

[Mark Goddard (mgoddard)]

FWIW, it should remove any IPs within the begin/end block.

[Radosław Piliszek (yoctozepto)]

... which users are unlikely to introduce for themselves.

[Mark Goddard (mgoddard)]

I just meant that if the IP changes and you run bootstrap-servers again, kolla-ansible should update the existing entry. It won't change anything it didn't introduce.

[Chason Chan (chen-xing)]

Should we remove the block of "ANSIBLE GENERATED HOSTS" or "the existing host entries"?

[Radosław Piliszek (yoctozepto)]

@Chason - the existing host entries

[Chason Chan (chen-xing)]

OK, will propose a fix. thanks@Radosław

[Mark Goddard (mgoddard)]

Would it always be safe to remove an entry that we did not add?

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/696047

[Radosław Piliszek (yoctozepto)]

Related: https://bugs.launchpad.net/kolla-ansible/+bug/1855935

[Radosław Piliszek (yoctozepto)]

Previously fixed loopback issue: https://bugs.launchpad.net/kolla-ansible/+bug/1837699