Closed tmccall8829 closed 4 months ago
@tmccall8829
As you can see from the error message:
| code: InvalidInput; message: Table orders
of schema
│ <redacted>
│ doesn't supports columns configuration
But you're trying to setup columns under this table:
"orders" = {
+ columns = {
+ "buyer_info_buyer_county" = {
In versions < 1.2.0 you receive the message from API side, in 1.2.0 we added client-side validation to terraform provider. Also - API didn't checked this as wel - we were accepting the request without error. Recently we fixed that.
The fact that you weren't facing this error below doesn't mean that you were able to configure columns. If you'll open Fivetran Dashboard in browser and open the schema configuration tab for the connector - you'll see that you're not able to configure or even expand table columns for this table.
Sorry for such experience, but now everything works as expected.
If you need to have an ability to configure columns for this table - probably you'll have to file a support ticket on support.fivetran.com about that. It's not a Fivetran Provider issue.
Thanks for the reply @beevital, but, as you'll see below, I don't agree that this issue should be closed.
I chatted with our Fivetran support rep, and that gave us a little bit more clarity:
orders
of schema ... doesn't support columns configuration" error, that's not because the connector itself doesn't support these columns; it's because fivetran hasn't pulled the schema yet for this connectorfivetran_connector_schema_config
resource actually useful for new connectorsFor example, just this morning, after the connector had been syncing all weekend, I successfully applied this configuration and restricted columns from this connector's schema.
So to me, that makes it seem like there should at least be some clearer error messaging around this. In the fivetran console/UI, for example, it shows this for newly created connectors:
My question is, why does the fivetran provider even support column schema configuration if you can't apply it to a connector without syncing some data first? That, to me, is quite a broken workflow, especially in the absence of a more meaningful error returned from the provider. It seems like it should at least be mirroring what the console says. If what our support rep is true (and it seems to be, based on what the fivetran console/UI is saying), then wouldn't this workflow be broken for any new connector created with a column-restriction schema config? It seems like this provider is attempting to provide functionality -- column schema restriction -- that the actual fivetran internals do not support.
Describe the bug Previously, we were defining fivetran connector schema configs using the
schema
arg. We created a number of amazon_selling_partner connectors that way, but the performance was very slow.Recently, we made the change to use
schemas
for better performance (as recommended by you all). We were able to successfully apply changes to existing connectors and their schemas using this new config. However, creating new connectors fails, with the following error message:To Reproduce Here is a snippet of how we're defining our resources:
Connector
Connector schedule
Connector schema config
Expected behavior Given that this definition matches what the current docs say to use, AFAICT, I'd expect the schema change to be applied properly/without issues.
Logs & Output Apply output is included above. Plan output:
Plugin version: v1.2.0, but noticed on v1.1.26 as well.
Additional context I've also filed a support case with fivetran, as it seems like this could also be related to this connector type specifically. But it's just odd that we were able to retroactively apply the new schema config to old connectors, while it fails for new ones.