Ah, forgot to mention that table is INNODB.


On Mon, Dec 30, 2013 at 1:59 PM, Cabbar Duzayak <cabbar@gmail.com> wrote:

> Hi,
>
>
> We have a basic table, which is something like :
>
>
> (id varchar50, productId varchar50, category varchar50)
>
>
> In this table, ID is the primary key and we have a unique index on
> (category, productId). And, there is a case where we want to do bulk
> inserts (3000 inserts) using:
>
>
> INSERT INTO (id, productId, category) VALUES (?x1?, ?y1?, ?z1?), (?x2?,
> ?y2?, ?z3?) ?.. ON DUPLICATE KEY productId = VALUES(productId), category > VALUES(category)
>
>
> So, when we try to insert 3000 rows using this syntax with a single
> statement, it takes ~ 3 seconds to execute this on an empty table.
>
>
> BTW, innodb_flush_log_at_trx_commit is set to 2 for us.
>
>
> I don?t have something factual data here, but it feels that this is too
> much just for 3000 rows.
>
>
> Isn?t this just too slow? Are we doing something wrong here? Or, can you
> advice as to which configuration parameters can we tweak here to make it
> faster?
>
>
> Thanks
>