I am using STCS compaction strategy for my 2 node Scylla Cluster. Each have 500G memory with a replication factor of 2 pointed to each other. The problem I am facing is that in STCS, SSTable size keeps increasing. So when there is only less than 100G memory remaining, the compaction doesn't complete because it runs out of disk space. So to avoid this from happening again, I want to reduce the maximum size of the SSTable and also, I want to do compaction more frequently so that less memory is used for the same.
I tried changing the strategy to LCS but then it tries to create 160mb SSTables for all the existing SSTables and again runs out of memory.
Please tell me how to change the above mentioned configurations. I am also open to any other better idea.