What do droppable tombstones mean in sstablemetadata for Cassandra?

It’s a bit confusing because estimated droppable tombstones can be larger than 1, so what does it mean? Does it mean that more than 100% of the data can be tombstoned, how is that possible?

Well, it doesn’t mean that. It means that for every 1 data cell, there’s x number of tombstone cells. So, if we have 4.0 droppable tombstones, then what we have is a ratio of non-tombstoned to tombstoned cells, 1:4. For every 4 tombstones, there’s 1 non-tombstoned cell. Or 25% of the table is non-tombstoned.

Here’s an example:

$ sstablemetadata ../../keyspace1/names-8752d4d1444b11ea886061d56f0fef73/ac-2150-bti-Data.db
SSTable: /var/lib/cassandra/data/keyspace1/names-8752d4d1444b11ea886061d56f0fef73/ac-2150-bti
Partitioner: org.apache.cassandra.dht.Murmur3Partitioner
Bloom Filter FP chance: 0.01
Minimum timestamp: 1582123535115908 (02/19/2020 14:45:35)
Maximum timestamp: 1585691070473353 (03/31/2020 21:44:30)
SSTable min local deletion time: 2147483647 (no tombstones)
SSTable max local deletion time: 2147483647 (no tombstones)
Compressor: org.apache.cassandra.io.compress.LZ4Compressor
Compression ratio: 0.4435361970714844
TTL min: 0
TTL max: 0
First token: -9223347087192471474 (544844)
Last token: 9223364646594345147 (837848)
minClusteringValues: [ lacerda ]
maxClusteringValues: [two]
Estimated droppable tombstones: 1.0

Here we have a 1:1 ratio, 1 data cell to 1 tombstoned cell.