Add metrics to count requests buffered to temporary file

[frittentheke]

Currently NGINX logs a warning for requests that are bigger than the configured client in-memory buffer (https://kubernetes.github.io/ingress-nginx/user-guide/nginx-configuration/configmap/#client-body-buffer-size):

[warn] 128#128: *5837 a client request body is buffered to a temporary file /tmp/nginx/client-body/0000000097, client: 10.xx.xx.xx, server: xxxx, request: "POST /api/v1/endpoint HTTP/2.0", host: "xxx", referrer: "h [...]

While logging this per request is already really helpful to find potential optimizations, having these events counted as a Prometheus metric would be even more helpful, as this is usually used by users and allows for correlation (e.g. buffered vs. non-buffered requests, ...).

[k8s-ci-robot]

This issue is currently awaiting triage.

If Ingress contributors determines this is a relevant issue, they will accept it by applying the triage/accepted label and provide further guidance.

The triage/accepted label can be added by org members by writing /triage accepted in a comment.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

[longwuyuan]

There are no resources like developer time available to add new features. Apologies for that. The reasons have been announced and available for listening on the recorded community meetings posted to youtube.

On a different note, if there is optional enabling of such a feature, we could consider a PR. But the real problem here is that you make no comments on the problems it will create. There will be too much data. In technical terms I saw someone use the words like "cardinality explosion", to refer to such situations.