Needlessly gathering facts - Performance issue

[pemsith]

SUMMARY

Since by default Ansible plays gather facts from all hosts, if you are not using any facts in tasks gather_facts should be set to false. There are multiple scripts where ansible facts are gathered but not used, which will be a performance issue.

ISSUE TYPE

• Bug Report

COMPONENT NAME

Performance

ANSIBLE VERSION

ansible [core 2.15.8]

COLLECTION VERSION

latest

OS / ENVIRONMENT

Cent OS 9

STEPS TO REPRODUCE

go to the below scripts and check. there are plays that gather facts but dont use.

tools/prepare_release.yml
tests/integration/targets/vmware_vm_inventory/prepare_environment.yml
tests/integration/targets/vmware_vm_inventory/test_inventory.yml

EXPECTED RESULTS

set gather_facts to false if facts aren't used.

[mariolenz]

This issue looks somewhat weird.

ANSIBLE VERSION

ansible [core 2.15.8]

We've removed support for ansible-core < 2.17 in 5.0.0. 4.x supported ansible-core 2.15, but it's EOL (#2507).

tools/prepare_release.yml

I am doing the releases here, and I don't see a problem there. A minute more, a minute less... who cares 🤷 Why do you want to run this?

tests/integration/targets/vmware_vm_inventory/prepare_environment.yml
tests/integration/targets/vmware_vm_inventory/test_inventory.yml

Do you really run those integration tests and run into any performance issues there?

[pemsith]

We have a tool that detects anti-patterns in Ansible, and I'm trying that out with selected OSS repos to see how accurate it is with real datasets.

One of the anti-patterns it detects is 'needless gathering of facts' (gathering Ansible facts even though it's not used). Your feedback is valuable for us to identify issues in our own tool so that we can further improve and provide the tool back to the community for use.

I understand that the first instance is a true positive, but the last 2 instances are related to integration tests and probably can be ignored, even though they match the anti pattern

[ihumster]

AFAIK gather_facts action plugin is part of ansible.builtin collection and his behavior is regulated in ansible.cfg or playbook for any play. Any of the community collections cannot influence the behavior of running ansible, but only provide ways to interact with managed hosts.

[mariolenz]

@pemsith Well, I think that your tool did find something interesting. I think I'll have a look at improving tools/prepare_release.yml. I don't want to touch the integration tests until we have a CI capable of running integrations tests again, though.

Just to make this clear, your findings don't affect users of the collection. They are "only" on stuff that's about (mostly internal) development processes of the collection.

Feel free to let us know if you think you found any more anti-patterns!

[mariolenz]

@pemsith BTW I wonder why your tool says that tools/prepare_release.yml is needlessly gathering facts, but ignores tools/unset_version.yml. I think gathering facts is unnecessary there, too.

[pemsith]

@mariolenz I really appreciate the feedback.

you are correct, 'tools/unset_version.yml' should have been caught, but it didn't. So that is a false negative. I see the issue and I will fix my tool to catch those cases in the future.