hachikuji commented on a change in pull request #9756:
URL: https://github.com/apache/kafka/pull/9756#discussion_r555466107
##########
File path:
raft/src/main/java/org/apache/kafka/raft/internals/BatchAccumulator.java
##########
@@ -37,6 +37,7 @@
private final Time time;
private final SimpleTimer lingerTimer;
private final int lingerMs;
+ private final int minFlushSize;
Review comment:
Batching has a lot of implications in the system. The flush behavior is
just one. It also impacts the segment size and the ability to do
down-conversion efficiently since we have to read the whole batch into memory.
Basically any time we need to do anything in memory with the batches, we have
to allocate at least enough memory to hold the largest batch. The controller is
designed to write small messages, so I do not think 1MB (say) would be much of
a constraint. For example, we have avoided storing assignment state in a single
message as we did with Zookeeper. We can reconsider it if ever needed, but I'd
like to keep batches relatively small if possible.
----------------------------------------------------------------
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:
us...@infra.apache.org
