yuqi1129 opened a new issue, #14498: URL: https://github.com/apache/gravitino/issues/14498
### Version main branch ### Describe what's wrong MCP REST clients inherit HTTPX's five-second read timeout. Valid metadata operations can take longer, and FastMCP can replace the upstream read-timeout context with a generic timeout. There is no CLI/environment setting for the Gravitino response read timeout. ### Error message and/or stacktrace A normal metadata tool call fails with a read timeout after five seconds. The MCP caller cannot tell which upstream timed out or whether a mutation may still complete. ### How to reproduce 1. Use an MCP tool with a Gravitino endpoint that takes more than five seconds to produce response data. 2. Invoke it through a real FastMCP client. 3. Observe the fixed read timeout and generic error. ### Additional context Make the read timeout finite/positive and configurable for every identity path; preserve authentication errors and never replay writes automatically. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
