How do I run top and ps commands inside middlware pod on OCP?

Issue

• How do I identify high CPU utilization by Java threads on OCP?

• Container images such as jboss-webserver-5/jws56-openjdk11-openshift-rhel8 and jboss-eap-7/eap74-openjdk11-openshift-rhel8, which are based on ubi8, do not have procps-ng package as follows:

  [lab@master ~]$ oc rsh <mw-pod-name>
  sh-4.4$ top -b -n 1 -H -p <java-pid>
  sh: top: command not found
  sh-4.4$ ps
  sh: ps: command not found
  

• CPU metrics such as container_cpu_user_seconds_total and container_cpu_system_seconds_total as cAdvisor metrics are collected by OpenShift Monitoring, but these metrics are in units of containers, and no cpu-usage metrics per thread in cAdvisor.

• Scraping interval of cpu utilization that can be referred in OpenShift monitoring stack is 30 seconds, but I need to check cpu utilization at finer intervals to identify burst load for a couple of seconds.
• Is Ephemeral Containers supported in OpenShift?