ChenSammi edited a comment on pull request #2694: URL: https://github.com/apache/ozone/pull/2694#issuecomment-938425499
> > > @ChenSammi Thanks for working on this! Please find my comments below. > > 1. Guava cache may not provide us the consistency guarantees we need for stateMachineData cache. For instance we must make sure that cache entry is evicted only after majority servers have written the data. But by limiting the number of entries in the guava cache any future addition of entries would lead to eviction of older entries. Ref: https://issues.apache.org/jira/browse/HDDS-2542 - We had seen a race condition earlier between read and write stateMachineData > > 2. The expiry time of 15mins after access might lead to retention of chunks for a longer time. This could increase the memory footprint of datanode. @lokeshj1703 , thanks for the feedback. Didn't notice that GuavaCache is used before HDDS-2542. With read/writeStateMachineData race condition resolved by HDDS-5619, it's a good time to rethink if GuavaCache is safe to use now. Guava cache with expire time will increae the memory footprint, this is a real cost when using Guava cache, with the benefit of keep the data in cache as long as possbile, that's the key point to reduce the slow follower happen. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
