During performance testing of 3scale 2.8 (https://issues.redhat.com/browse/THREESCALE-4878) it was detectred that currently the default resource requests/limts values of SystemAPP containers for CPU and memory are insufficient for large configurations.

Current default resource requests/limits:

• system-master:

   "limits": {"cpu": "1", "memory": "800Mi"}, "requests": {"cpu": "50m", "memory": "600Mi"}} 

   

• system-provider:

   "limits": {"cpu": "1", "memory": "800Mi"}, "requests": {"cpu": "50m", "memory": "600Mi"}} 

• system-developer:

   "limits": {"cpu": "1", "memory": "800Mi"}, "requests": {"cpu": "50m", "memory": "600Mi"}} 

• system-sidekiq:

   "limits": {"cpu": "1", "memory": "2Gi"}, "requests": {"cpu": "100m", "memory": "500Mi"}} 

• system-sphinx:

   "limits": {"cpu": "1", "memory": "512Mi"}, "requests": {"cpu": "80m", "memory": "250Mi"}} 

Trying to provision 3scale with:

    1 Account
        10000 Applications
    100 products
        1 application plan per product
        10 application plan limits per product
        100 application per plan
        10 backend usages per product
    1000 backend (each product will be using 10 backends)
        50 methods
        50 mapping rules

The provisioning process fails, 3scale API returning timeout errors.

If resources limits are disabled, the provisioning finishes successfully.

There is an ongoing work to expose custom resource limits https://issues.redhat.com/browse/THREESCALE-2507

So, this issue is about deciding if default request/limits values have to be changed. And update them if applies.