3 ms·
Yeah. The USB case is a cop out. For USB, keeping the pages dirty and the fsyncs erroring (as seems consistent with Postgres' needs and common sense) seems fi
by loeg 9y ago
Yeah. The USB case is a cop out. For USB, keeping the pages dirty and the fsyncs erroring (as seems consistent with Postgres' needs and common sense) seems fine.
The memory can be reclaimed when 'umount --force', or something like that, discards filesystem dirty state.