Journal.scanJournal reuses the same ByteBuffers for each entry, but (if the entry is large enough) EntryMemTable retains a reference. The result is that if any entry is large enough, the bytes that get written to the entryLog upon flush will be a prefix of the last entry replayed from that journal file. This will in general corrupt any such entries. Fix seems relatively simple, PR will follow.
[ Full content available at: https://github.com/apache/bookkeeper/issues/1737 ] This message was relayed via gitbox.apache.org for [email protected]
