ableegoldman commented on a change in pull request #10788:
URL: https://github.com/apache/kafka/pull/10788#discussion_r683102378
##########
File path:
streams/src/main/java/org/apache/kafka/streams/processor/internals/TopologyMetadata.java
##########
@@ -53,44 +56,162 @@
private static final Pattern EMPTY_ZERO_LENGTH_PATTERN =
Pattern.compile("");
private final StreamsConfig config;
- private final SortedMap<String, InternalTopologyBuilder> builders; // Keep
sorted by topology name for readability
+ private final TopologyVersion version;
+
+ private final ConcurrentNavigableMap<String, InternalTopologyBuilder>
builders; // Keep sorted by topology name for readability
Review comment:
Well I removed the whole `assignedNamedTopologies` thing from this PR,
but yes, that was one of the original intentions of adding it to the
AssignmentInfo. I'm leaning towards keeping things simple for now and avoid
adding on to the AssignmentInfo where possible. We can revisit this if the
extra rebalances become a problem.
It could be sufficient to just use the set of actually-assigned topologies
as a proxy. It's not perfect since it could be that the leader was aware of a
topology and simply happened to not assign any of its tasks to this client, but
I'd bet this gets rid of most of the unnecessary rebalances.
That said, I'm saving this optimization for a followup Pt. 4 PR. I'll
revisit the best way to handle the `builders` map in that as well
--
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.
To unsubscribe, e-mail: jira-unsubscr...@kafka.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
