Our requests are extremely inaccurate compared to what the container can/will use. From what I've observed, each container pegs a CPU at pretty much 100%, and RAM usage far exceeds 2Gi.

The most immediate thing we can do to help prevent bulk analysis from toppling nodes is to increase the requests for both of these.

WHAT WE NEED TO DOCUMENT:

Mainly we just need to highlight how a user can change the values in the operator and advise customers to implement eviction thresholds to protect their nodes:

• Linking to documentation and advising customers to implement eviction thresholds to protect their nodes.
  https://docs.openshift.com/container-platform/4.12/nodes/nodes/nodes-nodes-resources-configuring.html#allocate-eviction-thresholds_nodes-nodes-resources-configuring

OTHER SOURCE OF INFORMATION

https://issues.redhat.com/browse/MTA-123: see this ticket for log files showing the issue

https://github.com/konveyor/tackle2-operator/pull/176: see this PR for the proposed solution