Description
2 issues found, both will be hard to reproduce to verify fixes. Both issues occurred when ingesting large amounts of data (200GB).
1. https://issues.apache.org/jira/browse/ASTERIXDB-1710 Recovery redo fails whenever deadlock-free locking protocol invoked
Redo fails on WAIT log record – On restart, there was one log record type that prevented ingestion to continue, causing a hang. Will be hard to QE to reproduce.
2. Fix a mutual waiting condition when switching the txn log file:
– LogManager waits for LogBuffer to flush,
– LogBuffer is not full and cannot flush.