[GitHub] [activemq-artemis] wy96f commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate journal/page/large message file (AGAIN)

[GitBox <gi...@apache.org>]

wy96f commented on issue #2845: ARTEMIS-2336 Use zero copy to replicate journal/page/large message file (AGAIN)
URL: https://github.com/apache/activemq-artemis/pull/2845#issuecomment-541494236
 
 
   @franz1981
   
   > fix https://github.com/netty/netty/blob/ff7a9fa091a8bf2e10020f83fc4df1c44098bbbb/transport/src/main/java/io/netty/channel/DefaultMessageSizeEstimator.java#L45 in order to account for chunked files correctly and flow control them while being enqueued into chunk writer
   
   Not sure. I think netty is just limiting data in memory so not accounting for chunked file and file region?
   
   > change how sync done timeout is being calculated and consider the total time to replicate (from the first file region/chunk sent) instead of the time from when the sync done is being enqueued in Netty
   
   We block all journal related operations during sync done interval to make sure replica has received all data, so we‘d better not change it.
   
   Maybe we can:
   
   * We use original implementation when broker doesn't support zero region. IMO the chunked file way is more or less same with original one. At some point original version might be better bcs it reads file data in non-netty thread that can be beneficial where netty threads is busy taking care of new sent packets.
   
   * Another possible alternative is to use rate limiter like guava instead of netty flow control to work around the problem?
   
   

----------------------------------------------------------------
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:
users@infra.apache.org


With regards,
Apache Git Services