From a4f1a1e843ad39670aa449d4c1260ae16326e649 Mon Sep 17 00:00:00 2001
From: asonix <asonix@asonix.dog>
Date: Thu, 25 Jan 2024 17:11:40 -0600
Subject: [PATCH] Add 0.5.4 release document

---
 releases/0.5.4.md | 33 +++++++++++++++++++++++++++++++++
 1 file changed, 33 insertions(+)
 create mode 100644 releases/0.5.4.md

diff --git a/releases/0.5.4.md b/releases/0.5.4.md
new file mode 100644
index 0000000..7728939
--- /dev/null
+++ b/releases/0.5.4.md
@@ -0,0 +1,33 @@
+# pict-rs 0.5.4
+
+## Overview
+
+Another quick release to improve postgres performance, and reduce growth of the job_queue during
+cleanup spikes.
+
+### Changes
+
+- [Limit LISTEN Connections](#limit-listen-connections)
+- [Limit Spawned Jobs](#limit-spawned-jobs)
+
+
+## Descriptions
+
+### Limit LISTEN Connections
+
+When postgres is configured, pict-rs sometimes issues LISTEN queries in order to be notified of new
+jobs in the job queue, due to connection pooling, it would be possible for every connection pict-rs
+makes to postgres to be listening for notifications. This is needless and wastes resources. 0.5.4
+now limits the number of connections issuing LISTEN queries to at most 4, or as few as the number of
+cores in your machine, if less than 4.
+
+
+### Limit Spawned Jobs
+
+As mentioned in the 0.5.2 release document, pict-rs runs cleanup jobs that can spawn a lot of other
+jobs. In order to keep the queue size from growing too large for larger deployments, pict-rs now
+limits each general cleanup job to 2048 iterations, meaning it will complete after spawning just
+2048 jobs.
+
+Since pict-rs processes jobs in-order, this means that the majority of these jobs will be completed
+before the next general cleanup job runs, keeping the maximum number of queued jobs around 2048.