Open theoryxu opened 3 months ago
@diqiu50 Hi, will this issue also be fixed in 0.5.0?
Yes
@theoryxu
I attempted to start our gravitino-trino-connector on a worker, but it was unsuccessful. The reason is that our current catalog injection mechanism relies on finding the catalogChanger instance within the Trino ConnectorContext object, which cannot be located in worker mode.
After my testing, if the worker is configured to operate in a coordinator+worker mode, it can function properly. Although these worker nodes are configured with a coordinator, in reality, it does not work, because the discovery.uri
configured on all workers points to the actual coordinator.
An alternative solution would require modifying the Trino source code, which would complicate the deployment process.
Version
main branch
Describe what's wrong
When the node is a worker rather than a coordinator, there are some error and the node can't start.
Error message and/or stacktrace
How to reproduce
0.4.0
Additional context
related to https://github.com/datastrato/gravitino/issues/2376