I’ve long suspected that we need to put a delay before the ACK since we’re moving too damn fast for LDMP.
This is due the the problem described here:
"We get a bunch of records all at once, each one is published to NATS followed by an ACK to LDMP. After an arbitrary number of receipt/ACK pairs, the records stop coming in. Later, the health check detects the issue, then initiates a disconnect and reconnect cycle."
I’ve long suspected that we need to put a delay before the ACK since we’re moving too damn fast for LDMP.
This is due the the problem described here:
"We get a bunch of records all at once, each one is published to NATS followed by an ACK to LDMP. After an arbitrary number of receipt/ACK pairs, the records stop coming in. Later, the health check detects the issue, then initiates a disconnect and reconnect cycle."