Message-ID: From: "helloJosh (@helloJosh)" To: "pgjdbc/pgjdbc" Date: Tue, 25 Nov 2025 13:23:03 +0000 Subject: Re: [pgjdbc/pgjdbc] PR #3870: feat: Add INFO log when fetchSize is ignored under autocommit=true In-Reply-To: References: List-Id: X-GitHub-Author-Login: helloJosh X-GitHub-Comment-Id: 3575640558 X-GitHub-Comment-Type: issue_comment X-GitHub-Issue: 3870 X-GitHub-Repo: pgjdbc/pgjdbc X-GitHub-Type: comment X-GitHub-Url: https://github.com/pgjdbc/pgjdbc/pull/3870#issuecomment-3575640558 Content-Type: text/plain; charset=utf-8 Thanks for the helpful feedback. I understand your concern now - there are far more queries that return only a small number of rows than those that actually cause OOM. For option(b), I also think choosing a meaningful row-count is difficult, since too many rows are subjective and depends on local memory. Following your idea in (a), what do you think about the following behavior for a property `stabilityIssues=ignore|warn|fail` - When set to ignore (the default), pgjdbc would not do anything (current behavior). - When set to warn, pgjdbc would log a warning when a stability issue is detected (for example, fetchSize > 0 together with autocommit=true). - When set to fail, pgjdbc would throw an exception (or possibly log at a higher severity), so users can immediately see that a problematic pattern is being used.