spark-acid
spark-acid copied to clipboard
Revisit read Txn and Lock table for read operations
Currently read does not acquire locks due to life time of RDD is unspecified hence would lead to blocking of the InsertOverwrites. So protecting it against the DDL / Truncate / Compaction is operational issue as it stands.
Currently whenever DF materialises first time the transactional snapshot of read is acquired at that time.
Above two need rethinking ..
https://github.com/qubole/spark-acid/commit/2e66f76b3a3c974a73f03d9e80d9c66a0475a552
This commit adds frame work to perform operation under transactions.