Re: Is it really difficult for postgres_fdw to implement READCOMMITTED isolation? - Mailing list pgsql-hackers

From Ashutosh Bapat
Subject Re: Is it really difficult for postgres_fdw to implement READCOMMITTED isolation?
Date
Msg-id CAExHW5vdzthKWDJMMeJewLYM-O4VWYUKNJGQPfddae03vEuJxA@mail.gmail.com
Whole thread Raw
In response to Is it really difficult for postgres_fdw to implement READ COMMITTED isolation?  (Jinhua Luo <luajit.io@gmail.com>)
List pgsql-hackers

On Tue, Sep 18, 2018 at 8:28 PM Jinhua Luo <luajit.io@gmail.com> wrote:
https://www.postgresql.org/docs/current/static/postgres-fdw.html#id-1.11.7.43.12

As the doc said, the REPEATABLE READ isolation level is used to get
snapshot-consistent results.

But is it possible that postgres_fdw could get to know which remote
queries involved by each top outer command in the local transaction,
and use the same snapshot in the remote server to execute them
sequentially? For example, could we use PREPARE TRANSACTION and SET
TRANSACTION SNAPSHOT to archive this goal? Then we could use READ
COMMITTED on both sides?


I guess the problem is 1. exporting snapshots is not cheap 2. tracking prepared transactions is not implemented. See a nearby thread on "transaction involving multiple foreign server".

--
Best Wishes,
Ashutosh Bapat

pgsql-hackers by date:

Previous
From: Laurenz Albe
Date:
Subject: Re: pgsql: Allow concurrent-safe open() and fopen() in frontendcode for Wi
Next
From: Dilip Kumar
Date:
Subject: Re: Is it possible for postgres_fdw to push down queries onco-located tables?