franz1981 edited a comment 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-535443555 @wy96f Good analysis :) Yep, so the simplest possible solution I see is to tune differently `highWaterMark` > 1 MB to allow the chunk writer queue to continue to be drained. Both ChunkedInput and FileRegions are missing (on Netty) a correct size estimation on ChannelOutboundBuffer and this would imply senders to push many of them in burst: what makes them behave differently is that FileRegion is getting backpressured only by TCP while ChunkedInputs start to getting backpressured by Netty itself *into* ChunkedWriterHandler, given that any read ByteBuf is being accounted into ChannelOutboundBuffer thus preventing subsequent pending writes to proceed due to the small high watermark limit (if compared to the chunk size).
---------------------------------------------------------------- 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 With regards, Apache Git Services