Feature Request - Soft fail when tracer cannot load to prevent taking down Kubernetes ingress-nginx

[pzimno]

Related ingress-nginx issue

Currently raising NGX_ERROR seems to result in ingress-nginx going down. This could possibly be resolved by config validation in the ingress-nginx project or a different mechanism for handling errors in nginx-opentracing.

[kjivan]

Just wanted to mention we hit this yesterday and it was surprising to realize essentially our cluster broke because Jaeger went down.

[sriemer]

At IBM Observability by Instana APM we handle this with a K8s autotrace mutating webhook admission controller. It automatically does the required transformations of ConfigMap and deployment. Further components provide the required tracer binaries and symlinks to them fully automatically all together so that this case cannot happen.

We have the right timing and delay the startup of the NGINX of Ingress NGINX until the tracer is configured properly.
This solves a chicken-and-egg problem.