Hi @collimarco
I don’t have a LB running currently, so can’t answer, but since this isn’t documented in the various tutorials, then it would be nice with an official answer - can you help @jtittle ?

@collimarco

I’ve put in a request for more information so I can get a definite answer for you :).

My initial thought was that it was a request queue or QoS, though I spun up 3x Droplets with NGINX and an LB, then hit it with about 100k requests over 60 seconds (using the standard index file) and it didn’t budge.

HAProxy does have request queuing and QoS in place, though that’s only configurable if you deploy it on your own, so I’ll have to follow up once I hear back.

@collimarco

I apologize for the delay!

This is the number of connections that are queued up but not sent. So if you see a spike here it’s likely that your backend needs to have more servers or that there’s an issue with the servers.

What that means is that if everything is running smoothly, you won’t see that number jump. If you are seeing a spike on that column, it means that you may need to add additional servers to the LB so that it can efficiently distribute the load, or there are problems elsewhere, such as on the Droplets.

If you have any other questions, feel free to tag me and let me know!