fresh-borzoni opened a new issue, #2996: URL: https://github.com/apache/fluss/issues/2996
### Search before asking - [x] I searched in the [issues](https://github.com/apache/fluss/issues) and found nothing similar. ### Fluss version 0.9.0 (latest release) ### Please describe the bug 🐞 Filesystem config passed via `client.fs.*` prefix (introduced in #938) is silently ignored by all filesystem plugins. Two issues: 1. `FlussConnection` uses `extractPrefix()` instead of `extractAndRemovePrefix()`, keeping the `client.fs.` prefix intact. No plugin recognizes keys starting with `client.fs.`. 2. Even after fixing (1), OSS and OBS plugins pass config keys through to Hadoop as-is without rewriting to the Hadoop format. For example, `oss.endpoint` stays `oss.endpoint` but Hadoop expects `fs.oss.endpoint`. S3, Azure, and GS plugins correctly rewrite (e.g., `s3.endpoint` -> `fs.s3a.endpoint`), but OSS and OBS don't. **Example:** User sets `client.fs.s3.path-style-access = true` in catalog config. It's accepted without error but silently ignored. S3 client uses virtual-hosted style, job hangs or fails with DNS errors against S3-compatible stores. ### Solution `FlussConnection`: use `extractAndRemovePrefix()` to strip the `client.fs.` prefix Add short prefixes (`oss.`, `obs.`, `azure.`) to plugin prefix lists Make OSS and OBS plugins rewrite keys to Hadoop format, consistent with S3/Azure/GS ### Are you willing to submit a PR? - [x] I'm willing to submit a PR! -- 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]
