Also, clients that can't afford to wait for the broker when it goes down
can run embedded brokers that will queue published messages without
blocking the client. But there are disadvantages to using embedded brokers,
so be sure you understand the tradeoffs before you go down that path.
Tim
On
Which version are you using? Normally on start up the broker should
not need to replay the entire journal if the index already exists. On
startup the broker tries to determine the last in doubt position from
the index and only replay from that point. With HA I would expect
this to work the same