rangareddy commented on issue #17256: URL: https://github.com/apache/hudi/issues/17256#issuecomment-5508894958
One narrowing detail for whoever picks this up, since the report identifies the failing case only as `[3]`. `@EnumSource` iterates the enum in declaration order and JUnit 5 numbers parameterised invocations from 1, so the indices map onto `HoodieCDCSupplementalLoggingMode` as declared: | case | mode | | --- | --- | | `[1]` | `OP_KEY_ONLY` | | `[2]` | `DATA_BEFORE` | | `[3]` | **`DATA_BEFORE_AFTER`** | So the flake is specific to `DATA_BEFORE_AFTER`, the mode that keeps both before and after images in the supplemental log, and not to CDC streaming generally. That is the one mode where the reader can build the change details straight from the log rather than inferring an image, so it exercises a different code path from the other two. The test itself is unchanged and still running, at `TestCDCStreamingSuite.scala:45-47`, with no `@Disabled` in the file. This does not resolve anything: as noted earlier, the CDC read path was substantially rewritten after this was filed (#13444 / HUDI-9526 and #13694 / HUDI-9695), so the flake may well be gone, but a rewrite is not evidence for a specific race and I have no CI history to draw a flake rate from. Anyone with access to recent CI runs can settle it quickly now that the case is pinned to a single mode; if it has not recurred since those merges, this can be closed. -- 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]
