add gather facts on all hosts in inventory

[LuckySB]

/kind bug
What this PR does / why we need it:
playbook failed with scale.yml

no facts gather for kube-masters group, but need in fallback_ip_base vars definition

Which issue(s) this PR fixes:
Fixes #5927

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: LuckySB

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [LuckySB]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[LuckySB]

/assign @mattymo

[mattymo]

This won't solve the related issue because you usually run scale.yml with --limit argument. As a result, you won't gather facts on all hosts here. We had previously a much worse task that used delegate_to to gather facts on all nodes. It slowed deployments tremendously and I had to remove it.

[LuckySB]

Then we need to refactor the assembly cycle for variable falback_ips_base

limit is evil

[EppO]

can you run the scale playbook without limit? what would it do on the other nodes?
Using a new variable to list the new nodes to add would help, no?

[fktkrt]

This won't solve the related issue because you usually run scale.yml with --limit argument. As a result, you won't gather facts on all hosts here. We had previously a much worse task that used delegate_to to gather facts on all nodes. It slowed deployments tremendously and I had to remove it.

Can using the setup module be a viable option to fine-tune this?

[mattymo]

@LuckySB Correct. For some reason the lookup in hostvars is failing for nodes without cache. That should be improved to fall back to 127.0.0.1

[fktkrt]

Added a comment regarding hostvars lookup.

[fktkrt]

Shouldn't this be ansible_default_ipv4.address?

Or using something like this:
when: (hostvars[item] is defined) and (hostvars[item]['ansible_default_ipv4'] is defined) and (hostvars[item]['ansible_default_ipv4']['address'] is defined)

[EppO]

Do we have a CI test for the scale playbook? I know we have one for the upgrade but couldn't find anything for scale up/down.

[LuckySB]

@LuckySB Correct. For some reason the lookup in hostvars is failing for nodes without cache. That should be improved to fall back to 127.0.0.1

final result will depend on --limit flag
It is not idempotent

[LuckySB]

Do we have a CI test for the scale playbook? I know we have one for the upgrade but couldn't find anything for scale up/down.

no

[champtar]

should we also add meta: clear_facts, so we are sure we don't have some old set_facts around ?

[LuckySB]

well.
try to transfer the fallback_ips_base variable assembly to
roles /kubespray-defaults/tasks
with delegate_to: localhost
and run_once: true.

and set_facts: fallback_ips: hostvar.localhost.fallback_ips

[LuckySB]

/hold

[champtar]

You need to ignore errors, else we can't run the playbook when a host is down