Re: Simple Insert - very slow

From: Andy Sayer <andysayer_at_gmail.com>
Date: Fri, 17 May 2019 18:28:36 +0100
Message-ID: <CACj1VR4rHGavEuVkyhDrFUE7EMcEQhf2Q4dqGUWqXvOa7SAK4w_at_mail.gmail.com>



First step would be to see which object is responsible for all the IO, in my experience it’s usually down to one index.

Have a look at
https://ctandrewsayer.wordpress.com/2017/02/16/not-all-indexes-are-created-equally/ for how I’ve tackled this before.

It might be you’ve got a “bad index” or it might be that you’ve got an okay index that turned “bad”.

We had a case where an index exploded and was purely ITL for the first few thousands leaf blocks, we reset the sequence (told to by support) which meant that all our inserts had to scan those few thousand leaf blocks to insert one row.

First step is identifying the object

Andy

On Fri, 17 May 2019 at 16:47, Sanjay Mishra <dmarc-noreply_at_freelists.org> wrote:

> Hi All
>
> We have a table with around a billion rows with one Primary Key. Number of
> columns in the table are fewer than 10 and all are numbers data type.
> Application is doing few million inserts but running very slow. It is
> Exadata with two node and 10CPU and good SGA (bufer Cache 10G). Waits are
> coming heavily only for "cell single block physical read" for Insert
> process. While doing SQL Monitor shows it is doing high IO like more than
> 100G for simple Insert. Oracle version is 12.2.
>
> Any suggestions as what can be the reason for this wait event in this
> scenario.
>
> TIA
> Sanjay
>

--
http://www.freelists.org/webpage/oracle-l
Received on Fri May 17 2019 - 19:28:36 CEST

Original text of this message