We are frequently running out of resources across nodes due to high CPU/mem requests – but the actual utilization on the nodes is not that high. We need to optimize configs in the ephemeral cluster to ensure that CPU requests/limits and memory requests/limits are not set too high for apps. We could probably reduce nodes if we dial this in properly. The same processes we use to improve this on ephemeral can be useful for stage and production as well.

All apps need to determine how to set these values properly.

Devprod team should figure out how to best "analyze" the app to find the proper requests/limits, then document a best practice. We should then have each team follow the process to configure their app properly.

We can also modify bonfire to stop "stripping" resources. As long as every ClowdApp config uses a ${PARAMETER} for their resource limits/requests, and that parameter is set in app-interface under the ephemeral deploy target – we could decide that bonfire should "trust" that resource config.