You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@kafka.apache.org by GitBox <gi...@apache.org> on 2020/07/21 00:15:01 UTC

[GitHub] [kafka-site] ableegoldman commented on a change in pull request #275: MINOR: update Streams docs for 2.6 and fix configs

ableegoldman commented on a change in pull request #275:
URL: https://github.com/apache/kafka-site/pull/275#discussion_r457761691



##########
File path: 26/streams/upgrade-guide.html
##########
@@ -95,7 +95,19 @@ <h3><a id="streams_api_changes_260" href="#streams_api_changes_260">Streams API
         Note that you need brokers with version 2.5 or newer to use this feature.
     </p>
     <p>
-        As of 2.6.0 Kafka Streams deprecates <code>KStream.through()<code> if favor of the new <code>KStream.repartition()</code> operator
+        For more highly available stateful applications, we've modified the task assignment algorithm to delay the movement of stateful active tasks to instances
+        that aren't yet caught up with that task's state. Instead, to migrate a task from one instance to another (eg when scaling out),
+        Streams will assign a warmup replica to the target instance so it can begin restoring the state while the active task stays available on an instance
+        that already had the task. The instances warming up tasks will communicate their progress to the group so that, once ready, Streams can move active
+        tasks to their new owners in the background. Check out <a href="https://cwiki.apache.org/confluence/x/0i4lBg">KIP-441</a>
+        for full details, including several new configs for control over this new feature.
+    </p>
+    <p>
+        New end-to-end latency metrics have been added. These task-level metrics will be logged at the INFO level and report the min and max end-to-end latency of a record at the beginning/source node(s)
+        and end/terminal node(s) of a task. See <a href="https://cwiki.apache.org/confluence/x/gBkRCQ">KIP-613</a> for more information.
+    </p>
+    <p>
+        As of 2.6.0 Kafka Streams deprecates <code>KStream.through()</code> if favor of the new <code>KStream.repartition()</code> operator

Review comment:
       I didn't notice this in time to fix in AK, but we're missing the closing tag here so the code formatting is running loose. I'll add this on to an AK PR or something so it gets fixed for real




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org