pre-checks fail w/cinder and centos

[James]

Prechecks will fail when:

enable_cinder: "yes",
enable_cinder_backend_*: "no"
kolla_base_distro: "centos"

This occurs because centos8 no longer uses tgtd, and prevents doing a out-of-kolla backend config as a sole one.

[Radosław Piliszek (yoctozepto)]

Well, it uses LIO, so it's probably irrelevant that tgtd was dropped, no?

[Radosław Piliszek (yoctozepto)]

Could you paste the error message for posterity?

[James (james-o-benson)]

TASK [cinder : Checking at least one valid backend is enabled for Cinder] ********************************************************************************************************************************************************************
fatal: [x.x.x.x]: FAILED! => {"changed": false, "msg": "Please enable at least one backend when enabling Cinder"}

[James (james-o-benson)]

Correct, that it uses lio, but the problem is that because of that, the prechecks will fail fairly early on in cinder and not continue. I suggest we modify ansible to continue without a valid backend for cent images because of this, and fail on ubuntu for example.

[James (james-o-benson)]

per yoctozepto conversation. it is an issue with the deployment side, not the prechecks. enable_cinder_backend_lvm should be set to yes. But the deployment side should not try to get the tgtd docker image.

[Radosław Piliszek (yoctozepto)]

Need info on versions of images and kolla-ansible.

[James (james-o-benson)]

These are my versions and runtime options.
kolla-ansible 10.1.0
kolla_base_distro: "centos"
kolla_install_type: "source"
openstack_release: "ussuri"

[Launchpad Janitor (janitor)]

[Expired for kolla-ansible because there has been no activity for 60 days.]