cli-utils/examples/alphaTestExamples/inventoryNamespace.md

Demo: Inventory with Namespace

This demo shows that the namespace the inventory object is applied into will get applied first, so the inventory object will always have a namespace to be applied into.

First define a place to work:

DEMO_HOME=$(mktemp -d)

Alternatively, use

DEMO_HOME=~/hello

Establish the base

BASE=$DEMO_HOME/base
mkdir -p $BASE
OUTPUT=$DEMO_HOME/output
mkdir -p $OUTPUT
GREEN='\033[0;32m'
RED='\033[0;31m'
NC='\033[0m' # No Color

function expectedOutputLine() {
  if ! grep -q "$@" "$OUTPUT/status"; then
    echo -e "${RED}Error: output line not found${NC}"
    echo -e "${RED}Expected: $@${NC}"
    exit 1
  else
    echo -e "${GREEN}Success: output line found${NC}"
  fi
}

Create the "app"

Create the config yaml for a config map and a namespace: (cm-a, test-namespace).

cat <<EOF >$BASE/config-map-a.yaml
apiVersion: v1
kind: ConfigMap
metadata:
  name: cm-a
  namespace: test-namespace
  labels:
    name: test-config-map-label
EOF

cat <<EOF >$BASE/test-namespace.yaml
apiVersion: v1
kind: Namespace
metadata:
  name: test-namespace
EOF

Run end-to-end tests

The following requires installation of kind.

Delete any existing kind cluster and create a new one. By default the name of the cluster is "kind".

kind delete cluster
kind create cluster

Use the kapply init command to generate the inventory template. This contains the namespace and inventory id used by apply to create inventory objects.

kapply init --namespace=test-namespace $BASE | tee $OUTPUT/status
expectedOutputLine "namespace: test-namespace is used for inventory object"

Apply the "app" to the cluster. The test-namespace should be configured, and the config map should be created, and no resources should be pruned. The test-namespace is created first, so the following resources within the namespace (including the inventory object) will not fail.

kapply apply $BASE --reconcile-timeout=1m | tee $OUTPUT/status
expectedOutputLine "namespace/test-namespace apply successful"
expectedOutputLine "configmap/cm-a apply successful"
expectedOutputLine "apply result: 2 attempted, 2 successful, 0 skipped, 0 failed"
expectedOutputLine "reconcile result: 2 attempted, 2 successful, 0 skipped, 0 failed, 0 timed out"

# There should be only one inventory object
kubectl get cm -n test-namespace --selector='cli-utils.sigs.k8s.io/inventory-id' --no-headers | wc -l | tee $OUTPUT/status
expectedOutputLine "1"

# Capture the inventory object name for later testing
invName=$(kubectl get cm -n test-namespace --selector='cli-utils.sigs.k8s.io/inventory-id' --no-headers | awk '{print $1}')

# There should be one config map that is not the inventory object
kubectl get cm -n test-namespace --selector='name=test-config-map-label' --no-headers | wc -l | tee $OUTPUT/status
expectedOutputLine "1"

# ConfigMap cm-a had been created in the cluster
kubectl get configmap/cm-a -n test-namespace --no-headers | wc -l | tee $OUTPUT/status
expectedOutputLine "1"

Now delete the inventory namespace from the local config. Ensure that the subsequent apply does not prune this omitted namespace.

rm -f $BASE/test-namespace.yaml
kapply apply $BASE --reconcile-timeout=1m | tee $OUTPUT/status
expectedOutputLine "prune result: 1 attempted, 0 successful, 1 skipped, 0 failed"
expectedOutputLine "reconcile result: 1 attempted, 0 successful, 1 skipped, 0 failed, 0 timed out"

# Inventory namespace should still exist
kubectl get ns test-namespace --no-headers | wc -l | tee $OUTPUT/status

# Inventory object should still exist
kubectl get cm/${invName} -n test-namespace --no-headers | wc -l | tee $OUTPUT/status

# ConfigMap cm-a should still exist
kubectl get configmap/cm-a -n test-namespace --no-headers | wc -l | tee $OUTPUT/status
expectedOutputLine "1"