CNTRLPLANE-4795: Add etcd-density crashloop-flood step

ci-operator/step-registry/openshift-qe/etcd-density-crashloop-flood/OWNERS

@@ -0,0 +1 @@
+../OWNERS

ci-operator/step-registry/openshift-qe/etcd-density-crashloop-flood/openshift-qe-etcd-density-crashloop-flood-commands.sh

@@ -0,0 +1,58 @@
+#!/bin/bash
+set -o errexit
+set -o nounset
+set -o pipefail
+# Do not enable `set -x`: these scripts read and export Elasticsearch
+# credentials (ES_USERNAME/ES_PASSWORD) and construct an authenticated
+# ES_SERVER URL; tracing would leak them into the CI logs.
+oc projects
+oc version
+pushd /tmp
+
+ES_SECRETS_PATH=${ES_SECRETS_PATH:-/secret}
+ES_HOST="search-ocp-qe-perf-scale-test-elk-hcm7wtsqpxy7xogbu72bor4uve.us-east-1.es.amazonaws.com"
+ES_PASSWORD=$(cat "${ES_SECRETS_PATH}/password")
+ES_USERNAME=$(cat "${ES_SECRETS_PATH}/username")
+if [ -e "${ES_SECRETS_PATH}/host" ]; then
+    ES_HOST=$(cat "${ES_SECRETS_PATH}/host")
+fi
+UUID=$(uuidgen)
+
+REPO_URL="https://github.com/cloud-bulldozer/e2e-benchmarking"
+LATEST_TAG=$(git ls-remote --tags "${REPO_URL}.git" | awk -F'refs/tags/' '{print $2}' | grep -v '\^{}' | sort -V | tail -n1)
+TAG_OPTION="--branch $(if [ "$E2E_VERSION" == "default" ]; then echo "$LATEST_TAG"; else echo "$E2E_VERSION"; fi)"
+git clone $REPO_URL $TAG_OPTION --depth 1
+pushd e2e-benchmarking/workloads/kube-burner-ocp-wrapper
+
+export WORKLOAD="etcd-density crashloop-flood"
+# Patch the indexing call in run.sh so the fingerprint records "crashloop-flood"
+sed -i '/index\.sh/s|WORKLOAD="$WORKLOAD"|WORKLOAD="crashloop-flood"|' run.sh
+
+# Variant-specific flags
+EXTRA_FLAGS="--iterations=${CRASHLOOP_FLOOD_ITERATIONS} --crashloop-replicas=${CRASHLOOP_REPLICAS} --soak-duration=${SOAK_DURATION}"
+# Flags not handled by run.sh
+EXTRA_FLAGS+=" --gc-metrics=${GC_METRICS} --profile-type=${PROFILE_TYPE}"
+EXTRA_FLAGS+=" --metrics-profile=${METRICS_PROFILES}"
+if [[ "${ENABLE_LOCAL_INDEX}" == "true" ]]; then
+    EXTRA_FLAGS+=" --local-indexing"
+fi
+if [[ -n "${USER_METADATA}" ]]; then
+    echo "${USER_METADATA}" > user-metadata.yaml
+    EXTRA_FLAGS+=" --user-metadata=user-metadata.yaml"
+fi
+EXTRA_FLAGS+=" ${KB_FLAGS}"
+
+export ES_SERVER="https://$ES_USERNAME:$ES_PASSWORD@$ES_HOST"
+export EXTRA_FLAGS UUID ADDITIONAL_PARAMS
+
+set +o errexit
+./run.sh
+RUN_EXIT_CODE=$?
+set -o errexit
+
+METRICS_FOLDER="collected-metrics-${UUID}"
+if [[ -d ${METRICS_FOLDER} ]]; then
+    cp -r ${METRICS_FOLDER} "${ARTIFACT_DIR}/"
+fi
+
+exit ${RUN_EXIT_CODE}

ci-operator/step-registry/openshift-qe/etcd-density-crashloop-flood/openshift-qe-etcd-density-crashloop-flood-ref.metadata.json

@@ -0,0 +1,11 @@
+{
+	"path": "openshift-qe/etcd-density-crashloop-flood/openshift-qe-etcd-density-crashloop-flood-ref.yaml",
+	"owners": {
+		"approvers": [
+			"perfscale-ocp-approvers"
+		],
+		"reviewers": [
+			"perfscale-ocp-reviewers"
+		]
+	}
+}

ci-operator/step-registry/openshift-qe/etcd-density-crashloop-flood/openshift-qe-etcd-density-crashloop-flood-ref.yaml

@@ -0,0 +1,98 @@
+ref:
+  as: openshift-qe-etcd-density-crashloop-flood
+  from_image:
+    namespace: ci
+    name: ocp-qe-perfscale-ci
+    tag: latest
+  cli: latest
+  env:
+  - name: E2E_VERSION
+    default: "default"
+    documentation: |-
+      Override the e2e version
+  - name: KUBE_BURNER_VERSION
+    default: "default"
+    documentation: |-
+      Override the kube burner version
+  - name: CRASHLOOP_FLOOD_ITERATIONS
+    default: "20"
+    documentation: |-
+      Number of namespaces to create crashlooping pods in. Prefixed so it only
+      applies to this step and does not collide with the ITERATIONS consumed by
+      other kube-burner workloads (e.g. crd-scale) that may be co-scheduled in
+      the same job.
+  - name: CRASHLOOP_REPLICAS
+    default: "100"
+    documentation: |-
+      Crashlooping pods per iteration. Each pod runs "exit 1" with
+      restartPolicy: Always, so it emits roughly 6 events per crash cycle
+      (Scheduled, Pulling, Pulled, Created, Started, BackOff) continuously.
+  - name: SOAK_DURATION
+    default: "30m"
+    documentation: |-
+      How long to let events accumulate after the pods are created. This is the
+      measurement window: the pods are only there to keep producing events.
+  - name: KB_FLAGS
+    default: ""
+    documentation: |-
+      Additional flags for the kube-burner command. Reused across other kube-burner based workloads
+  - name: GC
+    default: "true"
+    documentation: |-
+      Default is true, which means clean up the pod/resource that kube-burner ocp created, you can set it to false to keep the resource
+  - name: GC_METRICS
+    default: "false"
+    documentation: |-
+      Default is false, which means skip collecting metrics, you can set it to true to collect metrics.
+  - name: PROFILE_TYPE
+    default: "regular"
+    documentation: |-
+      Metrics profile type: regular, reporting, or both
+  - name: METRICS_PROFILES
+    default: "metrics.yml,etcd-density-metrics.yml"
+    documentation: |-
+      Comma separated list of metrics profiles to use
+  - name: ENABLE_LOCAL_INDEX
+    default: "false"
+    documentation: |-
+      Trigger to enable local indexing
+  - name: ES_SECRETS_PATH
+    default: ""
+    documentation: |-
+      Override elasticsearch secrets path.
+  - name: USER_METADATA
+    default: ""
+    documentation: |-
+      Arbitrary metadata provided to the kube-burner workload in YAML format
+  - name: ADDITIONAL_PARAMS
+    default: '{}'
+    documentation: |-
+      JSON value to upload additonal metadata to the fingerprint
+  commands: openshift-qe-etcd-density-crashloop-flood-commands.sh
+  timeout: 13h
+  credentials:
+  - collection: ocp-perfscale-cpt
+    group: ocp-qe-perfscale-es-creds
+    mount_path: /secret
+  - collection: stackrox-perfscale-elasticsearch
+    group: credentials
+    mount_path: /secret_stackrox
+  resources:
+    requests:
+      cpu: 100m
+      memory: 100Mi
+  documentation: >-
+    This step runs the etcd-density crashloop-flood workload to stress etcd with
+    realistic event pressure from crashlooping pods.
+
+
+    Unlike event-storm, which creates synthetic Event objects, this deploys pods
+    that actually crash and so generates events from kubelet, scheduler and
+    controller sources at once. That exercises the count-increment (PATCH) path
+    for keys that already exist, a different etcd mvcc profile from creates, and
+    the shape most real-world event storms take.
+
+
+    Defaults are kube-burner-ocp's own. The OpenShift Performance and Scale team
+    characterised standard against sharded etcd with
+    CRASHLOOP_FLOOD_ITERATIONS=20 and SOAK_DURATION=10m on a 24 worker cluster.