e2e: test for leaving CPUs in busy shared pools.

Add a test case from Antti Kervinen <antti.kervinen@intel.com>
for testing that busy shared CPU pools are not sliced empty.

Co-authored-by: Antti Kervinen <antti.kervinen@intel.com>
Signed-off-by: Krisztian Litkey <krisztian.litkey@intel.com>

Author: klihub

test/e2e/policies.test-suite/topology-aware/n4c16/test15-busy-shared-pools/code.var.sh

@@ -0,0 +1,34 @@
+vm-command "kubectl delete pods --all --now"
+helm-terminate
+
+helm_config=$(COLOCATE_PODS=false instantiate helm-config.yaml) helm-launch topology-aware
+
+# Create 4 BestEffort containers, which should be spread across NUMA nodes,
+# putting one container in each. Then create a Guaranteed container with a
+# CPU request equal to the number of CPUs in a NUMA node. This container
+# cannot be put into any NUMA node, since then it would exhaus the shared
+# pool, which we can't do in any NUMA node as they are all busy with a
+# BestEffort container. So the Guaranteed container should be placed in a
+# socket, slicing off CPUs from both of its NUMA nodes.
+
+CONTCOUNT=4 create besteffort
+CPU=4 MEM=100M CONTCOUNT=1 create guaranteed
+verify 'disjoint_sets(cpus["pod0c0"],cpus["pod0c1"],cpus["pod0c2"],cpus["pod0c3"],cpus["pod1c0"])'
+
+vm-command "kubectl delete pods --all --now"
+
+# Repeat the same test using Burstable containers that look like BestEffort
+# based on their CPU request alone.
+CONTCOUNT=4 CPUREQ=1m CPULIM=750m create burstable
+CPU=4 MEM=100M CONTCOUNT=1 create guaranteed
+verify 'disjoint_sets(cpus["pod2c0"],cpus["pod2c1"],cpus["pod2c2"],cpus["pod2c3"],cpus["pod3c0"])'
+
+vm-command "kubectl delete pods --all --now"
+
+# Repeat the same test using non-zero CPU request Burstable containers.
+CONTCOUNT=4 CPUREQ=250m CPULIM=750m create burstable
+CPU=4 MEM=100M CONTCOUNT=1 create guaranteed
+verify 'disjoint_sets(cpus["pod4c0"],cpus["pod4c1"],cpus["pod4c2"],cpus["pod4c3"],cpus["pod5c0"])'
+
+vm-command "kubectl delete pods --all --now"
+helm-terminate