Re: IO related waits

On Fri, Sep 20, 2024 at 4:47 PM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:

> veem v <veema0000(at)gmail(dot)com> writes:
> > Able to reproduce this deadlock graph as below. Now my question is ,
> this
> > is a legitimate scenario in which the same ID can get inserted from
> > multiple sessions and in such cases it's expected to skip that (thus "On
> > conflict Do nothing" is used) row. But as we see it's breaking the code
> > with deadlock error during race conditions where a lot of parallel
> threads
> > are operating. So how should we handle this scenario?
>
> Do you have to batch multiple insertions into a single transaction?
> If so, can you arrange to order them consistently across transactions
> (eg, sort by primary key before inserting)?
>

That's exactly what I did back in the day. Because of database buffering,
sorting the data file at the OS level made the job 3x as fast as when the
input data was random.

--
Death to <Redacted>, and butter sauce.
Don't boil me, I'm still alive.
<Redacted> crustacean!