Hi Nikita,
So, this is targeting only the scenario with multiple producers and one consumer?
Mostly, yes. Because of the log format the transactions can't be read out in parallel.
(Or does GTID_EVENT or something else store the length of transaction in the log?)
I see SEQ_READ_APPEND is only used for Relay Log on the slave. Afaiu, the relay
log has only only one producer, the network io thread. Do I miss something
here? Please clarify.
Right, therefore only single-producer case can be applied to the replication.
We have one scenario on ES with multiple producers though.