terraform apply failed for offer.microceph

[Marian Gasparovic]

During cluster bootstrap

terraform apply failed: data.juju_offer.microceph[0]: Reading...
data.juju_offer.microceph[0]: Read complete after 0s [id=admin/controller.microceph]

https://oil-jenkins.canonical.com/job/fce_deploy_sunbeam/578//console

artifacts

https://oil-jenkins.canonical.com/artifacts/7d8903cf-6867-4eb5-b032-f0753ba0e09a/index.html

[James Page (james-page)]

The offer is read from the environment, not created by the plan so this feels like a bug in the terraform provider somewhere.

Error: cannot add saas application "microceph": saas application already exists

  with juju_integration.glance-to-ceph[0],
  on main.tf line 264, in resource "juju_integration" "glance-to-ceph":
 264: resource "juju_integration" "glance-to-ceph" {

The plan generates to relations to the offered SAAS - glance-to-ceph happens to be the second one created. That said this is very much an occasional issue rather than on all runs (and all single-node deploys in SolQA enable ceph).

[Hemanth Nakkina (hemanth-n)]

Is this during enabling/disabling telemetry plugin?

There is an issue with removing offers which has multiple consumers and is fixed in juju terraform provider 0.10.0 [1]

As part of the fix, any consume offer call checks if the saas application already exists and handles the case. Hopefully the fix should be good enough even during cluster bootstrap.

This bug need to be verified once all the terraform plans are updated to use juju terraform provider 0.10.0

[1] https://github.com/juju/terraform-provider-juju/issues/308
[2] https://github.com/juju/terraform-provider-juju/pull/309

[Hemanth Nakkina (hemanth-n)]

Sunbeam is using terraform-provider-juju 0.11.0 version now with the fixes mentioned in #2.

Marian, Are you seeing this problem anymore?