Question about TTL
Created by: m-attack
I've tried running 3 node cluster setup with TTL 1 hour but noticed that data doesn't actually get deleted. When I look over the logs, I see that vacuum processes are running every hour, but then I run query asking for deleted data and here it is.
Then I experimented with single-node setup to find out when the garbage was collected. I created bucket with PPF 1h and TTL 1h via ddb-admin and started load generator while monitoring disk usage, load averages, top, etc. After an hour disk usage didn't stop to grow, and in 12h DB ate all available space. After that frontend stopped responding on mps requests.
Later I tried combination of PPF 1m, TTL 1h with the same result, except that performance was worse due to the small amount of data per file.