1996fanrui commented on PR #27035: URL: https://github.com/apache/flink/pull/27035#issuecomment-3352279128
Thanks @rkhachatryan and @Zakelly for the comments! > 1. Are you sure that we don't disable auto-compaction for "normal" DB (i.e. options object is not reused)? I am sure `auto-compaction` is not disabled for "normal" DB. The reason is `restoreTempDBInstanceFromLocalState` only create temp rocksdb instance, and all of options object are closed while closing temp instance, these options are not shared/reused. > Should it be covered by a test? Good point, I have added a new `testAutoCompactionIsDisabled` to cover this case. FYI: I will squash test commit before merging. >> Should we replicate this for ForSt? > Yeah, we could do similar for ForSt in another PR. After looking into the `ForStIncrementalRestoreOperation#mergeStateHandlesWithClipAndIngest`, I do not think ForSt has `IndexOutOfBounds` bug. When `exportedColumnFamilyMetaData.isEmpty()`, it uses the original `restoreStateHandles` [1] fallback instead of `notImportableHandles`. `restoreStateHandles` never be empty here, but `notImportableHandles` may be empty, causing the bug. [1] https://github.com/apache/flink/blob/ba8312f05b4b8836ba658a78288f681b8a8ed76a/flink-state-backends/flink-statebackend-forst/src/main/java/org/apache/flink/state/forst/restore/ForStIncrementalRestoreOperation.java#L695 -- 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]
