merlimat opened a new pull request #8634:
URL: https://github.com/apache/pulsar/pull/8634


   ### Motivation
   
   The updates on the managed cursor are (by default) rate limited to avoid 
generating a lot of write operations when consumers are acknowledging messages.
   
   Due to the nature of the rate limiting, if the traffic on topic suddenly 
stops, the last updates to the cursor position are kept only in memory and they 
would be flushed only in 2 conditions, either: 
    1. The traffic resumes and the consumer acks again
    2. The topic is gracefully closed
   
   If the topic is not gracefully closed, the cursor will be rolled back to the 
last persisted position.
   
   That will trigger re-delivery of potentially very old messages. Instead we 
should have a time-bound period after which we ensure that all cursor updates 
are flushed on disk.
   


----------------------------------------------------------------
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


Reply via email to