always set schema caches from provider clients #33543
Conversation
Allow core to always use the global schema cache, so that providers without GetProviderSchemaOptional are not spun up repeatedly. Rather than conditionally setting the cache, we just conditionally use the cache in the client to work around providers without GetProviderSchemaOptional.
|
Reminder for the merging maintainer: if this is a user-visible change, please update the changelog on the appropriate release branch. |
| @@ -79,12 +79,14 @@ func (p *GRPCProvider) GetProviderSchema() (resp providers.GetProviderSchemaResp | |||
| defer p.mu.Unlock() | |||
|
|
|||
| // check the global cache if we can | |||
| if !p.Addr.IsZero() { | |||
| if !p.Addr.IsZero() && resp.ServerCapabilities.GetProviderSchemaOptional { | |||
There was a problem hiding this comment.
this breaks the caching. resp is the zero value, i.e. GetProviderSchemaOptional is always false.
There was a problem hiding this comment.
That is intentional. Caching via GetProviderSchemaOptional must be opt-in to maintain current behavior with existing provider releases.
|
I'm going to lock this pull request because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active contributions. |
The logic for skipping the cache was incorrect, causing core to spin up non-GetProviderSchemaOptional providers too often.
Rather than conditionally setting the cache based on
GetProviderSchemaOptional, we can conditionally use the cache in the client to work around providers withoutGetProviderSchemaOptional, which allows core to still access the schema from the global cache and prevent repeatedly spinning up providers which don't haveGetProviderSchemaOptionalset. The client for these providers will have the schema also stored in the client instance to prevent additionalGetProviderSchemacalls, and act as a flag to ensure thatGetProviderSchemahas been called at least once.