Continuing Reporting Stack dependency discussion from today's Tech Committee Call

Hi Everyone,

I want to continue the conversation about Zookeeper from today’s tech committee call.

Here’s a brief summary of the discussion:

  • OpenLMIS uses Consul for many features that are similar to Zookeeper

  • Druid and Kafka rely on Zookeeper for managing their clusters. Currently, we will be deploying single node clusters for each of these services.

  • I’m personally not sure if Zookeeper is a hard dependency

  • There’s a bigger issue here for OpenLMIS. There’s an understanding that OpenLMIS is vertically scalable, but has not yet been tested to be horizontally scalable. Dockerizing the reporting stack will create a number of docker containers that make it easier for standing up dev environments on single node clusters. However, production environments may not be best served by docker containers for Kafka and Druid (specifically) because they have inbuilt processes that let them manage their own clustering.

I don’t have any specific questions at this time, but want to have the discussion here on the forum for community contribution.

FYA: Josh and Peter