Nginx logs those errors " worker_connections are not enough"

[Patrick Desnoyers]

Using this : https://github.com/charmed-kubernetes/charm-kubeapi-load-balancer

2020/04/16 14:32:19 [alert] 4886#4886: *171498101 768 worker_connections are not enough while connecting to upstream, client: 10.XXX.X.17, server: _, request: "GET /api/v1/namespaces/site-production/secrets?fi
eldSelector=metadata.name ......

It would be nice to be able to increase the worker_connections in the Charm

Should I make a pr ? It doesn't look very difficult to add this option.
(It could require that we increase some limits too?)

[Patrick Desnoyers (patrickd75)]

As the cluster grows (It's not that big, around 11 worker nodes), we are starting to see this error more and more.

Any opinions ?

[Tim Van Steenburgh (tvansteenburgh)]

Hi @patrickd75, thanks for reporting this! If you're interested in submitting a PR, that would be great - we'll be happy to review it.

[Patrick Desnoyers (patrickd75)]

I will start working on it shortly.

I don't think I agree with classification "wishlist". Connection to the API is pretty important, no ? If the LB starts overloading on a small cluster, what does it do on a bigger one ?

[Tim Van Steenburgh (tvansteenburgh)]

Hey @patrickd75, thanks, and I see your point about the Importance. I was thinking of it as a "feature to add new worker_connections configuration", but yeah, it could be considered a bug.