README

This is an implementation of a spout that receives data from Scribe. It is an example of a "push spout", in that data is pushed to it rather than the spout pulling data from somewhere else.

This spout requires a fork of Scribe that allows for Zookeeper-based node discovery: https://github.com/traviscrawford/scribe

See http://groups.google.com/group/scribe-server/msg/c6d04c3b0ee25ddc for more details on how that works.

This is an unreliable spout that doesn't guarantee data processing. With some more work, it can be extended to be reliable – see the note in ScribeReceiver.java.

This spout *mostly* respects TOPOLOGY_MAX_SPOUT_PENDING. If the spout is deactivated (like as part of the topology shutdown), it will flush its internal queue of messages to the output collector (which may be more than TOPOLOGY_MAX_SPOUT_PENDING). This ensures that every message pushed to it is at least attempted to be processed. As explained above, with more work actual reliability could be added to the spout.