i just optimized my system for long term database/large dataset type loads.
so nearly 7 months in an i have yet to deal with a database maintenance issue.
a few key points in this is sizable and big l2arc, all writes are sync writes and thus goes through a slog device / ram until its flushed to the hdd’s in sequential writes thus reducing fragmentation of the data on drives.
ofc these features will not clean bad database structure, they will only help keep the utilization smooth and snappy, enough so that there should be enough other nodes with issues before i feel them and thus storj should in theory take care of it
thats my approach anyways…
if doing vacuuming on a db is safe… well thats a larger debate, which i don’t know a massive amount about, i did watch some lectures about it… in which it sounded a bit like… one cannot really avoid it at times, but it will remove stuff from the database and thus atleast in theory will never make the data safer or more stable.
but isn’t it always like that, performance vs safety
can’t have both
personally i like to compare vacuum to trim on ssd’s, it also helps performance, but will ruin your ability to restore recently deleted data, and tho not the same as vacuum it is fairly similar from a conceptual point of view.
i do use trim on my ssd’s tho
never really did do much db stuff, so took the hardware approach to the problem, instead of learning new stuff lol