Retrieving “tombstoned” records in Cassandra

Question

My question is very simple. Is it in any way possible to retrieve columns that have been marked tombstone before the GCGraceSeconds period expiry(default 10 days). If yes what would be the exact CQL query for that?

If I were to understand the deletion process the tombstones are marked on the MemTables and the SSTable being immutable waiting for compaction still has the the deleted data waiting for compaction. So before compaction occurs is there any way to read the tombstoned data from either the Memtable or SSTable?

Using CQL 3.0 on CQLSH command prompt & Cassandra 2.0.

Answer 1

You are right, when a tombstone is inserted it usually doesn't immediately delete the underlying data (unless all your data is in a memtable). However, you can't control when it does. If you don't have much data and compaction happens quickly, the underlying data may be deleted very quickly, much sooner than 10 days.

There is no query to read deleted data, but you can inspect all your SSTables with sstable2json to see if they contain the deleted data.

Answer 2

Just to add on to the previous comment. Have a low value of gc_grace_seconds for the column families that have frequent deletions. It will take some time for gc but tombstones are expected to get cleared .