We see that terraform detects drift from our LD instance because fallthrough.rollout_weights becomes []. To avoid the noise in detection, I think the provider could handle nil better instead of waiting for the reconciliation.
Thanks for bringing this to our attention. I've created an internal ticket to investigate a solution. We'll be sure to update this issue as we learn more.
Using provider version
2.0
We see that terraform detects drift from our LD instance because
fallthrough.rollout_weights
becomes[]
. To avoid the noise in detection, I think the provider could handlenil
better instead of waiting for the reconciliation.My best guess is that https://github.com/launchdarkly/terraform-provider-launchdarkly/blob/main/launchdarkly/fallthrough_helper.go#L73 could be enhanced to default to
[]
before the percent rollout check.