#!/bin/bash

set -euo pipefail

SCRIPTPATH="$( cd "$(dirname "$0")" ; pwd -P )"
PRESERVE=true

export TEST_WINDOWS="false"
export AEMM_VERSION="1.8.1"
export AEMM_DL_URL="https://github.com/aws/amazon-ec2-metadata-mock/releases/download/v$AEMM_VERSION/amazon-ec2-metadata-mock-$AEMM_VERSION.tgz"
export CLUSTER_CONFIG_FILE=$SCRIPTPATH/cluster-spec.yaml


USAGE=$(cat << 'EOM'
  Usage: run-test [-a script1,scripts2,...] [-w] [-d] CONFIG

    Options:
      -a       Assertion script(s), default is ALL scripts in e2e dir
      -w       Target Windows platform
      -d       Delete cluster upon test completion/failure

    Arguments:
      CONFIG   File to source, it should export the following values:
                 CLUSTER_NAME                                 EKS cluster name
                 DOCKER_PULL_POLICY                           Docker image pull policy (defaults to IfNotPresent)
                 NODE_TERMINATION_HANDLER_DOCKER_REPO         Node Termination Handler Docker repository
                 NODE_TERMINATION_HANDLER_DOCKER_TAG          Node Termination Handler Docker tag
                 NODE_TERMINATION_HANDLER_DOCKER_PULL_POLICY  Node Termination Handler Docker image pull policy (defaults to DOCKER_PULL_POLICY)
                 WEBHOOK_DOCKER_REPO                          Webhook Proxy Docker repository
                 WEBHOOK_DOCKER_TAG                           Webhook Proxy Docker tag
                 WEBHOOK_DOCKER_PULL_POLICY                   Webhook Proxy Docker image pull policy (defaults to DOCKER_PULL_POLICY)
                 EC2_METADATA_PORT                            EC2 Metadata Test Proxy port (defaults to 18999)
                 EC2_METADATA_URL                             EC2 Metadata Test Proxy URL (defaults to "http://amazon-ec2-metadata-mock-service.default.svc.cluster.local:$EC2_METADATA_PORT")
                 WEBHOOK_URL                                  Webhook URL (defaults to "http://webhook-test-proxy.default.svc.cluster.local")
EOM
)

while getopts "a:dw" opt; do
  case ${opt} in
    a ) # Assertion script(s)
        assertion_scripts=$(echo $OPTARG | tr "," "\n")
      ;;
    d ) # DELETE EKS Cluster + ECR repos
      echo "␡️ This run will delete the eks cluster and ecr repos"
      PRESERVE=false
      ;;
    w ) # Windows platform
        TEST_WINDOWS="true"
      ;;
    \? )
      echo "$USAGE" 1>&2
      exit
  esac
done

function exit_and_fail {
    local pod_id
    pod_id=$(get_nth_worker_pod || :)
    kubectl logs $pod_id --namespace kube-system || :
    assertion_end=$(date +%s)
    echo "⏰ Took $(expr $assertion_end - $assertion_start)sec"
    echo "❌ NTH EKS Integration Test FAILED $CLUSTER_NAME! ❌"
    exit 1
}

function retry {
  local retries=$1
  shift

  local count=0
  set +e
  trap "set -e" RETURN
  until "$@"; do
    exit=$?
    set -e
    wait=$((2 ** $count))
    count=$(($count + 1))
    if [ $count -lt $retries ]; then
      echo "Retry $count/$retries exited $exit, retrying in $wait seconds..."
      sleep $wait
    else
      echo "Retry $count/$retries exited $exit, no more retries left."
      return $exit
    fi
    set +e
  done
  return 0
}
export -f retry

function clean_up {
    if [[ "$PRESERVE" == false ]]; then
      echo "====================================================================================================="
      echo "🧹  Cleaning up EKS and ECR  🧹"
      echo "====================================================================================================="
      eksctl delete cluster -f $CLUSTER_CONFIG_FILE || :
      aws ecr --region $REGION delete-repository --repository-name $NTH_REPO_NAME --force >/dev/null 2>&1 || :
      aws ecr --region $REGION delete-repository --repository-name $WEBHOOK_REPO_NAME --force >/dev/null 2>&1 || :
      return
    fi
}

# Required to execute tests
deps=("docker" "eksctl" "helm" "kubectl" "jq")
for dep in "${deps[@]}"; do
    path_to_executable=$(which $dep)
    if [ ! -x "$path_to_executable" ]; then
        echoerr "You are required to have $dep installed on your system..."
        echoerr "Please install $dep and try again. "
        exit 3
    fi
done

TEST_OS="linux"
if [[ $TEST_WINDOWS = "true" ]]; then
  TEST_OS="windows"
fi

export TEST_WINDOWS \
  TEST_OS

config=${*:$OPTIND:1}
if [[ -z ${config} ]]; then
  echo "no CONFIG provided; creating EKS cluster using $CLUSTER_CONFIG_FILE"
  # shellcheck source=../eks-cluster-test/provision-cluster
  source $SCRIPTPATH/../eks-cluster-test/provision-cluster
else
  echo "Reading configuration from ${config}"
  set -a  # Export variables by default.
  # shellcheck disable=SC1090
  source $config
  set +a  # Disable exporting variables by default.
  PRESERVE=true # cluster user-provided so don't try to delete
fi

echo "CLUSTER_NAME=${CLUSTER_NAME:?"not found"}"
echo "DOCKER_PULL_POLICY=${DOCKER_PULL_POLICY:="IfNotPresent"}"
echo "NODE_TERMINATION_HANDLER_DOCKER_REPO=${NODE_TERMINATION_HANDLER_DOCKER_REPO:?"not found"}"
echo "NODE_TERMINATION_HANDLER_DOCKER_TAG=${NODE_TERMINATION_HANDLER_DOCKER_TAG:?"not found"}"
echo "NODE_TERMINATION_HANDLER_DOCKER_PULL_POLICY=${NODE_TERMINATION_HANDLER_DOCKER_PULL_POLICY:=$DOCKER_PULL_POLICY}"
echo "WEBHOOK_DOCKER_REPO=${WEBHOOK_DOCKER_REPO:?"not found"}"
echo "WEBHOOK_DOCKER_TAG=${WEBHOOK_DOCKER_TAG:?"not found"}"
echo "WEBHOOK_DOCKER_PULL_POLICY=${WEBHOOK_DOCKER_PULL_POLICY:=$DOCKER_PULL_POLICY}"
echo "EC2_METADATA_PORT=${EC2_METADATA_PORT:=18999}"
echo "EC2_METADATA_URL=${EC2_METADATA_URL:="http://amazon-ec2-metadata-mock-service.default.svc.cluster.local:$EC2_METADATA_PORT"}"
echo "WEBHOOK_URL=${WEBHOOK_URL:="http://webhook-test-proxy.default.svc.cluster.local"}"

# The e2e test scripts use other variable names.
echo "IMDS_PORT=${IMDS_PORT:=$EC2_METADATA_PORT}"
echo "INSTANCE_METADATA_URL=${INSTANCE_METADATA_URL:=$EC2_METADATA_URL}"

# Label first linux node in cluster; nth will be deployed to this node ONLY -- NEEDED FOR LINUX RUN ONLY:
if [[ $TEST_OS = "linux" ]]; then
  nth_node=$(kubectl get node --selector="kubernetes.io/os=linux" --output jsonpath='{.items[-1].metadata.name}')
  export NTH_WORKER_LABEL="role=nth_worker"
  kubectl label nodes $nth_node $NTH_WORKER_LABEL --overwrite
fi

export NODE_TERMINATION_HANDLER_DOCKER_PULL_POLICY \
    WEBHOOK_DOCKER_PULL_POLICY \
    WEBHOOK_URL \
    IMDS_PORT \
    INSTANCE_METADATA_URL

if ! command -v kubectl >/dev/null 2>&1; then
    echo "🚫 Cannot find kubectl command"
    exit 1
fi
if ! kubectl get svc >/dev/null 2>&1; then
    echo "🚫 kubectl failed test communication with cluster"
    exit 1
fi

TEST_NODE=$(kubectl get nodes -l kubernetes.io/os=$TEST_OS |
    tail -n +2 |
    tr -s "\t" " " |
    cut -d" " -f1)
export TEST_NODE
echo "TEST_NODE=${TEST_NODE:?"not found"}"

function get_nth_worker_pod {
    kubectl get pods -n kube-system -l k8s-app=aws-node-termination-handler |
        tail -n +2 |
        tr -s "\t" " " |
        cut -d " " -f 1 | head -1
}
export -f get_nth_worker_pod

# shellcheck disable=SC2120
function reset_cluster {
    echo "-------------------------------------------------------------------------------------------------"
    echo "🧹 Resetting cluster $CLUSTER_NAME"
    echo "-------------------------------------------------------------------------------------------------"
    $SCRIPTPATH/../eks-cluster-test/reset-cluster
    sleep ${1:-15}
}

if [[ -z ${assertion_scripts+x} ]]; then
    assertion_scripts=(
        "$SCRIPTPATH/../e2e/cordon-only-test"
        "$SCRIPTPATH/../e2e/imds-v2-test"
        "$SCRIPTPATH/../e2e/maintenance-event-cancellation-test"
        "$SCRIPTPATH/../e2e/maintenance-event-dry-run-test"
        #"$SCRIPTPATH/../e2e/maintenance-event-reboot-test"
        "$SCRIPTPATH/../e2e/maintenance-event-test"
        "$SCRIPTPATH/../e2e/spot-interruption-dry-run-test"
        "$SCRIPTPATH/../e2e/spot-interruption-test"
        #"$SCRIPTPATH/../e2e/webhook-http-proxy-test"
        #"$SCRIPTPATH/../e2e/webhook-secret-test"
        "$SCRIPTPATH/../e2e/webhook-test"
    )
fi

echo "Assertion script(s): ${assertion_scripts[*]}"

trap "exit_and_fail" INT TERM ERR
trap "clean_up" EXIT

for assertion_script in "${assertion_scripts[@]}"; do
    reset_cluster
    echo "================================================================================================="
    echo "🥑 Running assertion script $(basename $assertion_script)"
    echo "================================================================================================="
    assertion_start=$(date +%s)
    $assertion_script
    assertion_end=$(date +%s)
    echo "⏰ Took $(expr $assertion_end - $assertion_start)sec"
    pod_id=$(get_nth_worker_pod || :)
    kubectl logs $pod_id --namespace kube-system || :
    echo "✅ Assertion test $assertion_script PASSED! ✅"
done
reset_cluster

echo "====================================================================================================="
echo "✅ All tests passed! ✅"
echo "====================================================================================================="