You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@kafka.apache.org by ij...@apache.org on 2022/09/13 19:18:52 UTC

[kafka] branch 3.3 updated: MINOR: Add upgrade note regarding the Strictly Uniform Sticky Partitioner (KIP-794) (#12630)

This is an automated email from the ASF dual-hosted git repository.

ijuma pushed a commit to branch 3.3
in repository https://gitbox.apache.org/repos/asf/kafka.git


The following commit(s) were added to refs/heads/3.3 by this push:
     new 6c745b53bf1 MINOR: Add upgrade note regarding the Strictly Uniform Sticky Partitioner (KIP-794)  (#12630)
6c745b53bf1 is described below

commit 6c745b53bf1b66f2f650cb1d00d98be59658c0c0
Author: Artem Livshits <84...@users.noreply.github.com>
AuthorDate: Tue Sep 13 11:56:44 2022 -0700

    MINOR: Add upgrade note regarding the Strictly Uniform Sticky Partitioner (KIP-794)  (#12630)
    
    Reviewers: Ismael Juma <is...@juma.me.uk>, David Jacot <dj...@confluent.io>
---
 docs/upgrade.html | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/docs/upgrade.html b/docs/upgrade.html
index 276af1812a6..f6ad74e999d 100644
--- a/docs/upgrade.html
+++ b/docs/upgrade.html
@@ -81,6 +81,11 @@
 
 <h5><a id="upgrade_330_notable" href="#upgrade_330_notable">Notable changes in 3.3.0</a></h5>
     <ul>
+        <li>The partitioner used by default for records with no keys has been improved to avoid pathological behavior when one or more brokers are slow.
+            The new logic may affect the batching behavior, which can be tuned using the <code>batch.size</code> and/or <code>linger.ms</code> configuration settings.
+            The previous behavior can be restored by setting <code>partitioner.class=org.apache.kafka.clients.producer.internals.DefaultPartitioner</code>.
+            See <a href="https://cwiki.apache.org/confluence/display/KAFKA/KIP-794%3A+Strictly+Uniform+Sticky+Partitioner">KIP-794</a> for more details.
+        </li>
         <li>There is now a slightly different upgrade process for KRaft clusters than for ZK-based clusters, as described above.</li>
         <li>Introduced a new API <code>addMetricIfAbsent</code> to <code>Metrics</code> which would create a new Metric if not existing or return the same metric
             if already registered. Note that this behaviour is different from <code>addMetric</code> API which throws an <code>IllegalArgumentException</code> when