cliff@krumvieda.com wrote:
> Hi, Tom:
>
> >Whichever one manages to get to the index page first will go through.
> >The second one will block waiting to see if the first one commits,
> >and will error out if so --- or proceed, if it aborts.
>
> I see, this makes sense. What if the two transactions insert rows
> that don't violate the constraint: will they be able to proceed in
> parallel? Or will one wait for the other, because they both need to
> update the shared index? I.e., does the mechanism work by waiting for
> one index update to commit before permitting the next?
They proceed in parallel. Backends wait only on specific rows that
conflict, not the index page.
--
Bruce Momjian <bruce@momjian.us> http://momjian.us
EnterpriseDB http://postgres.enterprisedb.com
+ If your life is a hard drive, Christ can be your backup. +
