Message-ID: <gh-pgjdbc-pgjdbc-194-c1149192044@github.com>
From: "rbygrave (@rbygrave)" <noreply+rbygrave@github.com>
To: "pgjdbc/pgjdbc" <noreply+pgjdbc-pgjdbc@github.com>
Date: Tue, 07 Jun 2022 21:35:11 +0000
Subject: Re: [pgjdbc/pgjdbc] issue #194: PgJDBC can experience client/server deadlocks during batch execution
In-Reply-To: <gh-pgjdbc-pgjdbc-194@github.com>
References: <gh-pgjdbc-pgjdbc-194@github.com>
List-Id: <gh-pgjdbc-pgjdbc.github.com>
X-GitHub-Author-Login: rbygrave
X-GitHub-Comment-Id: 1149192044
X-GitHub-Comment-Type: issue_comment
X-GitHub-Edited-At: 2022-06-07T22:29:13Z
X-GitHub-Issue: 194
X-GitHub-Repo: pgjdbc/pgjdbc
X-GitHub-Type: comment
X-GitHub-Url: https://github.com/pgjdbc/pgjdbc/issues/194#issuecomment-1149192044
Content-Type: text/plain; charset=utf-8

> quite big but not crazy big - 20k rows, 10 columns

To me 20k sounds very big.  Have you benchmarked that against smaller batch sizes (like batch size 1000, 100, 50, 20)?  

When I benchmark test insert batching looking to find a reasonably optimal batch size I don't recall going over 1k (and generally it is way smaller batch size).  I kind of think of it as approximately filling a network packet, often 100 is a good batch size but it will depend on width of columns etc.

Edit: Read the full history and the test case - very interesting re NOTIFY and size of returning content and blowing out the buffers.