What happens when Indexer is down and UF is forwarding data without using Acknowledgement?

[jagadeeshm]

I have a multi-site indexer clustering.

All my UF(s) are configured for Site0 (auto-balanced across all indexers available in both sites) and Indexer Acknowledgement is DISABLED.

When UF tries to forward data to a specific indexer and if it is down, does UF re-try sending the SAME data to the next indexer available in the auto-load balancing list?

PS: I referred this document already and it is not clear enough about this specific situation -
http://docs.splunk.com/Documentation/Splunk/6.4.4/Forwarding/Protectagainstlossofin-flightdata

[woodcock]

Yes. The s2s (Splunk-to-Splunk) protocol handles this without data loss even without ACK.

[jagadeeshm]

any supporting documentation on this ?

[mattymo]

Two of the example troubleshooting forwarder topics elude to this:

https://docs.splunk.com/Documentation/Forwarder/6.5.2/Forwarder/Troubleshoottheuniversalforwarder

Essentially, a forwarder opens a tcp socket toward the indexers in the LB list. if it cant connect it moves to the other. You can see this in action by grepping splunkd.log | grep TcpOutputProc.

Once connected to an idx the uf will transmit data. If the UF cannot connect to any receiver, it fills its queues then stops ingesting data, picking up when there is a valid receiver to send to.

It is not exactly "in-flight data" as the forwarder needs to connect to a receiver before it will transmit