Hive Integration

[ardatan]

• Implementation of Hive plugin
• Implementation of Hive transform to track sources
• Implementation of Hive store
• Documentation

Motivation

Today Mesh CLI generates artifacts for each source and saves them on the disk. This has two disadvantages;

• Everytime source changes, Mesh needs to rebuild and deploy.
• Hard to track breaking changes on the source level.

If Mesh uses the CDN directly instead of the disk, Mesh doesn't need to be rebuilt and deploy. A quick restart should do the trick. Also a hot reload logic can be implemented since no redeployment is needed.

In the documentation we need to cover 2 different scenarios;

First scenario: Publish on source repos then fetch on the gateway
Mesh CLI needs to be installed on the source repo then Mesh needs to be configured with the specific handler. Then mesh publish will publish the generated schema/artifacts to Hive CDN. Then on the gateway;

openapi:
  source: http://hive-cdn-url/schema
  schemaHeaders:
    TOKEN-HEADER: TOKEN-VALUE

Second scenario: Publish on the gateway
mesh publish command will generate artifacts and publish the artifacts to Hive CDN instead of saving them on the disk so deployed Mesh code won't have anything about the source schemas. So the gateway configuration will remain the same, we should only add a flag or sth else to point Mesh to the Hive CDN in order to fetch artifacts on runtime.
Hive will basically replace FsStoreStorageAdapter.

[gilgardosh]

Implementation of Hive store: added as a standalone task to roadmap