Gather pod information to check IRQ load balancing

Some intensive CPU workload could be negatively affected by the
interruption of ISRs.
We would like to check that IRQ affinities are properly configured to
avoid problems.
That means no exclusively assigned CPU is listed in any IRQ affinity
list.

In kubernetes version >= 1.23 PodResourcesAPI provides us with the
exclusively assigned CPU for each container but in previous versions it
gives us all the CPUs used by the container which in fact could be the
default cpuset.
So for kubernetes version < 1.23 we need extra information to get the
exclusively assigned CPUs.
That is the list of pods with `qosClass == Guaranteed` so we could then
look for the CPUs in the PodResourcesAPI.

As the way to calculate exclusively assigned CPUs for kubernetes version
< 1.23 also works for kubernetes versions >=1.23, and it is a better way
to provide a homogeneous way to do it, at least until kubernetes
versions < 1.23 go out of support, we have decided to use this way for
both cases.

We fulfill the information gap we need with the bare minimum amount of
pod information we need to do the calculation, that is:
- pod namespace
- pod name
- pod qosClass

and we gather this information on a per node basis to make it easy to
cross-reference this information with the PodResourcesAPI info which is
also on a per node basis.

Author: jlojosnegros

must-gather/collection-scripts/gather_nodes

@@ -69,6 +69,8 @@ do
     oc exec $pod -n perf-node-gather -- gather_sysinfo --json podres --socket-path=unix:///host/podresources/kubelet.sock > $NODE_PATH/podresources.json
 
     oc exec $pod -n perf-node-gather -- gather_sysinfo snapshot --debug --root=/host --output=- > $NODE_PATH/sysinfo.tgz 2> $NODE_PATH/sysinfo.log
+
+    oc get pods -A --field-selector spec.nodeName=$node,status.phase=Running -o go-template='[{{range $idx, $item := .items}} {{if (ne $idx 0)}},{{end}}{"namespace":"{{.metadata.namespace}}", "name":"{{.metadata.name}}", "nodeName":"{{.spec.nodeName}}", "qosClass": "{{.status.qosClass}}" }{{"\n"}}{{end}}]' > $NODE_PATH/pods_info.json
 done
 
 # Collect journal logs for specified units for all nodes