So, for quick tests with the pulsar-standalone there's no need to worry about ZooKeeper, since it's providing everything out of the box.
So, you need to deploy a ZK cluster or connect to an existing one only if you're trying to deploy a multi-node Pulsar cluster, with "real" brokers and bookies. >From the perspective of Pulsar and Storm there is no requirement to share the same ZooKeeper cluster, though nothing prevents from doing it. I would say, that if the load and data set size on an existing ZK cluster are not that high, and you don't plan to create 100Ks of topics, there should be no problem in sharing the cluster. I guess the main challenges can arise if there are multiple teams interacting on the same ZK cluster with many different workloads. Some unexpected mistakes from one team could cause instabilities on the whole cluster. In practice, in a correctly setup hardware (multiple disks, enough mem and no swap) the ZK cluster can sustain very intensive workloads with super-high reliability. > Not too clear on how to get started with pulsar feeding into storm spout. Thanks I would just start with running the Pulsar standalone in one machine. Then follow the examples at https://pulsar.incubator.apache.org/docs/latest/adaptors/PulsarStorm/ for how to write a simple topology using the PulsarSpout adapter. You then just need to point the "serviceUrl" to the address of the machine where the Pulsar standalone service is running. On Wed, Aug 9, 2017 at 11:23 AM J.R. Pauley <[email protected]> wrote: > Hi. Thanks. Do you happen to know when using with Storm and trying out > PulsarSpout and Bolt, given that storm has its own zookeeper, what is best > practice for that? Do you want both storm and pulsar to have their separate > zookeeper, or share one with storm? If the latter where do I tell pulsar > not to start its own zookeeper. Not too clear on how to get started with > pulsar feeding into storm spout. Thanks > -- Matteo Merli <[email protected]>
