Pros and Cons of push or pull ingestion?


Just a general question about different types of how Druid ingest data. Currently, we’re using pull data which we use RabbitMQ. But we have seen a lot of people using tranquility which we would like to know the pros and cons?

We’re thinking of rolling Stream Pull from Amazon Kinesis but have seen people doing that with Tranquility.

Sorry that I didn’t finish the writing.

From what I’ve read so far. Tranquility is like a bunch of servers reading data from producer (e.g Storm or Kafka) and pushing to Druid. What would happen if one of the server failed? Would it lose the data?

I guess this can be achieved with the same pull mode having a lot of realtime nodes pulling data?

I’m just trying to figure out which works the best in terms of HA?

Thanks in advance.

The limitations of the stream pull model is highlighted here:

RabbitMQ may not have the same limitations as Kafka.