You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Jan 24, 2024. It is now read-only.
Describe the bug
Currently, KoP handles FetchRequest by reading all requsted partition in one round. If topic has lots of partitions, e.g. 1000, the response will be very huge. Which will makes the channel un-writablle.
To Reproduce
Steps to reproduce the behavior:
Start pulsr with kop enabled
Create a partitioned topic with lots of partitions, e.g. 1000
Send messages to this topic from kafka client
Consume message from this topic
Expected behavior
KoP should handle this circumstances, even with many partitions, consume should be work normal.
Screenshots
When consumer we can see channel un-writable
From the view of client, consume hangs for a period of time, consume may be continue or not.
Additional context
Even we disable the logic of configure channel auto read when channelWritabilityChanged like this
@OverridepublicvoidchannelWritabilityChanged(ChannelHandlerContextctx) throwsException {
if (log.isDebugEnabled()) {
log.debug("Channel writability has changed to: {}", ctx.channel().isWritable());
}
/* if (ctx.channel().isWritable()) { // set auto read to true if channel is writable. ctx.channel().config().setAutoRead(true); } else { log.debug("channel is not writable, disable auto reading for back pressing"); ctx.channel().config().setAutoRead(false); ctx.flush(); } ctx.fireChannelWritabilityChanged();*/
}
the problem still happens
The text was updated successfully, but these errors were encountered:
Describe the bug
Currently, KoP handles
FetchRequest
by reading all requsted partition in one round. If topic has lots of partitions, e.g. 1000, the response will be very huge. Which will makes the channel un-writablle.To Reproduce
Steps to reproduce the behavior:
Expected behavior
KoP should handle this circumstances, even with many partitions, consume should be work normal.
Screenshots
When consumer we can see channel un-writable
From the view of client, consume hangs for a period of time, consume may be continue or not.
Additional context
Even we disable the logic of configure channel auto read when channelWritabilityChanged like this
the problem still happens
The text was updated successfully, but these errors were encountered: