I started to witness some non expected behavior with FDB in one of our cluster:
The database is fairly large: 1TB and we have a total of 6 replica (3 for the main region and 3 for the second) and we end up using close to ~12TB ( I guess this is because of write amplification due to the use of the SQLlite engine) but when I did a backup yesterday I saw the total disk space growing by around 1.5TB (ie. more or less the size of the DB) when the write rate is around 2MB/s (or 7GB per hour). The backup lasted ~ 3 hours so it shouldn’t have contribute to more than 1TB of total disk space even when you take into account the amplification due the number of replicas and the amplification due to the use of the SQLlite backend.
I’m wondering if someone had ever witnessed that.