Proposed title of this feature request

Loki - Allow Using Resource Requests as Limits

What is the nature and description of the request?

There are customers who, for architectural reasons, need to give Loki resource limits.  This Feature would allow the existing Loki resource requests to also be used as limits, so that they are not only used for scheduling but also enforced during runtime of Loki.

Why does the customer need this? (List the business requirements)

If a customer is on-prem, they may be on a Bare Metal cluster, and their available resources for that cluster are constrained.  If Loki loses connectivity with storage, Loki will continue to retain logs on the cluster, and if Loki cannot send those logs to storage, there is a risk that Loki would consume all the resources on the node before those logs are offloaded to storage.  This could cause the node to fail.

If optional Loki "resource limits" are enabled, and Loki loses storage connectivity, then Loki would drop logs once its resource limits are reached.  At this point, Loki will run out of resources, may restart, and logs will be lost.   However, the Loki resources would not consume the entire node.

List any affected packages or components.

Loki Operator

Non-Goals:

This RFE will not allow supported user-configurable resource limits for Loki that differ from those listed in the documented Loki t-shirt sizes.