Indexer Acknowledgement and stalling of data flow

[tomrit]

Hi!

I have a setup where I must clone and forward data to a third party. Can somebody clarify if I disable useACK that even though a destination is unreachable that the flow to other outputs does not stall? If I look at the spec of outputs.conf, for me it's not fully clear what happens when I write to the socket and the destination is unreachable:

* When set to "false", the forwarder considers the data fully processed
  when it finishes writing it to the network socket.

Thanks in advance!

[tomrit]

Thanks a lot for the fast clarification!

[acharlieh · Accepted solution]

Even with useAck disabled, if one side is completely blocked, then Splunk cannot send so the output Queue could back up and can eventually cause sending to all outputs to halt.

(on a forwarder with useAck true the forwarder  removes the data from its buffer after it gets acknowledgement that the indexer finished processing the data. With useAck false, the forwarder removes the data from the buffer after the indexer successfully received the data… if the forwarder cannot connect in either case, it fills the buffer)

I think you’re looking for something like setting

dropEventsOnQueueFull

On the third party output, others may have more experience with tuning this and similar settings ref: 

https://docs.splunk.com/Documentation/Splunk/8.2.3/Admin/Outputsconf