Pulsar retries on failures. So if you want Pulsar to do the retries, you can send the timeout to be a higher value. If you want to be able to control the retry behavior, you can low the timeout value. The timeout value should be larger than your normal publish time latency. The publish time latency is usually dominated by the disk latency.
Hope this provides some guidance for you. - Sijie On Fri, Jul 23, 2021 at 8:04 AM Marco Guest <marcogues...@gmail.com> wrote: > > Hi, > my company will use Pulsar as messaging system for a geographically > distribuited architecture on Kubernetes. > In our use case, an application using C++ Pulsar client will send up to 300 > 000 message/sec on twenty Pulsar permanent topics by asynchronous send. > C++ Producer has a default send timeout value of 30 seconds to receive > acknowledge from the broker. It seems to us a too much big timeout before to > raise up an error. > How can we evaluate a value for a more strict control but sufficient for our > use case? > > Thanks > >
