Implement Round Robin/Load balancing

[ph]

On the first payload of events, the plugin will randomly choose a server and will keep that connection open until something bad happen to the server. It would be great if we had the option to round robin between sending batches.

[ph]

cc @jakommo

[lsoumille]

Any progress on this ?

[GazzaC]

I would like to hear news on this item also. I have a tiered implementation of logstash which is connected via the lumberjack protocol and the logstash-output-lumberjack and logstash-input-lumberjack plugins. I also employ the logstash multi pipeline feature and connect multi-piplines vua lumberjack on a 2 tiered arrangement. Are there any plans on this happening soon?

[ydewinte]

Any update?

[systemdave]

Has there been any further discussion on a path forward for this functionality Logstash to Logstash?

[tomerilp]

any chance to add this capability?

[tfrederick74656]

I am also very interested in this functionality. Working with an implementation ow that consists of larger individual Logstash nodes at various remote sites that aggregate to a scale-out set of lower-spec cloud-hosted Logstash nodes. It's impossible to achieve horizontal scalability at the final aggregation point when the downstream nodes use sticky node selection and may be larger than any individual node in the upstream set.

[nextsummer33]

Just checking the code in lumberjack client, A host is selected by shuffling the array and pop the last. As the result, it just create a single connection rather than multiple connections and flush the data to each host by the round robin.