Scott Hess wrote:
> WRT #3, you could also consider journal_mode off (or memory, if your code
> requires transactions to work right). In that case, the database state is
> indeterminate if you have an app-level crash, but you should be fine if you
> make it to the end.
It would be a better idea
What I'd do:
1) Look at the indices, and make sure the input is sorted to insert in
index order. Also drop any unnecessary indices and add them back at the
end. [Read the code for vacuum to see what kinds of things make sense to
defer.]
2) Bump up the cache a lot. Inserting in sorted order
Do you have some indices on the table? Drop them before the huge insert and
then recreate again. It is faster.
On Mon, 24 Apr 2017 11:00:00 -0700
Jens Alfke wrote:
> I’m importing a large data set with a lot of rows — an entire Wikipedia dump,
> about 60GB, one article per
3 matches
Mail list logo