lhotari commented on PR #15748:
URL: https://github.com/apache/pulsar/pull/15748#issuecomment-1145288916

   > Can you elaborate on what kind of regression it was seen there? I have in 
general experienced the opposite, going from msg-count based to memory based.
   
   We have run into regressions with pulsar-perf that seem to be caused by the 
change in the memory limit. Some test results are very sensitive of the 
producer's `maxPendingMessages` and `maxPendingMessagesAcrossPartitions` 
settings (as well as `batchingMaxBytes`, `batchingMaxPublishDelayMicros` and 
`batchingMaxMessages` settings). For a testing tool, I think it's unexpected 
that the parameters wouldn't have an impact in the CLI tools such as 
`pulsar-perf`. That's why I think it's better to disable the memory limit until 
there's a configurable parameter.
   
   > I believe a better change here is to make the memory limit configurable 
for the perf producer.
   
   I agree. Until that is done, it's better to set the limit to 0 by default. I 
added #15912 for tracking this.


-- 
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: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to