Duplicate layer of AppendBatchSizeTrackers #5531
Labels
area/client
kind/enhancement
Improvements that should be made
priority/P2
version/0.9.0
version/0.10.0
Problem description
The writer uses
AppendBatchSizeTracker
objects to infer the right size of batches to be written to the server. However, currently there are 2 places in which suchAppendBatchSizeTracker
objects are being instantiated:FlowHandler
andTcpClientConnection
. In principle, the batching logic should be only in one place, and concretely, at the connection level. This is because, without connection pooling enabled, he number of flows should be equal to the number of connections. With connection pooling enabled, is at the connection level at where we need to track the data being written to the server, which can be a mix of multiple multiplexed flows.In fact, after a preliminary analysis, it looks like the
AppendBatchSizeTracker
objects instantiated inFlowHandler
are not used at all, so we can safely remove them.Problem location
Unnecessary
AppendBatchSizeTracker
objects instantiated inFlowHandler
.Suggestions for an improvement
Remove unnecessary code.
The text was updated successfully, but these errors were encountered: