Connection Pooling from a single app is pretty easy and straight forward but what is the best way to manage connection pool for a storm topology. Since every bolt may potentially run in a separate node & JVM, creating a connection pool for each container could result in many open connections which could be an issue.
Is there a way to share connection pool for a topology? Are there any best practices / lessons learns, anyone would like to share?
I would use tick tuple to write to RDBMs as you can easily overwhelm database otherwise. Here's a great article on best practices http://hortonworks.com/blog/apache-storm-design-pattern-micro-batching/