#!/usr/bin/env bash

set -euo pipefail

ROOT=$(cd $(dirname ${BASH_SOURCE[0]}) && pwd)
source ${ROOT}/lib.sh

remote_config=${CONFIG_PATH}
local_config=/tmp/base-config.yaml
bucket_name=${remote_config#s3://}
bucket_name=${bucket_name%%/*}
output_bucket_region=$(aws s3api get-bucket-location --bucket=$bucket_name --query="LocationConstraint" \
  --region=us-west-2 --output=text --no-cli-pager)
aws s3 cp ${remote_config} ${local_config}  --region=$output_bucket_region

o11y_repo=tidbcloud/observability
git_ref="main"
region=$(yq '.region' "${local_config}")
logical_region_name=$(yq '.name' "${local_config}")
env=$(yq '.env' "${local_config}")
vendor=$(yq '.cloudProvider' "${local_config}")
cluster=$(yq '.k8s.name' "${local_config}")
jobId=$(yq '.jobId' "${local_config}")
output=$(yq '.outputPath' "${local_config}")
account=$(yq '.aws.controlplaneAccount' "${local_config}")
imageRegistry=$(yq '.imageRegistry' "${local_config}")
regionalServerRoleARN="arn:aws:iam::${account}:role/regional-server-${cluster}"
hosted_zone_name="observability-$env.pingcap.cloud"
if [ "$env" == "prod" ]; then
    hosted_zone_name="observability.tidbcloud.com"
fi
echo "hosted_zone_name: $hosted_zone_name"

outputs=$(aws cloudformation describe-stacks \
  --stack-name auto-deploy-$cluster-ingress \
  --region="$region" \
  --query 'Stacks[0].Outputs' \
  --output json \
  --no-cli-pager)
certificateARN=$(echo $outputs | jq -r '.[] | select(.OutputKey=="RsCertificateArn") | .OutputValue')
wafACLARN=$(echo $outputs | jq -r '.[] | select(.OutputKey=="RsWebAclArn") | .OutputValue')
echo "certificateARN: $certificateARN"
echo "wafACLARN: $wafACLARN"

# remote_iam_config=$(yq '.jobOutput.iam' "${local_config}")
# local_iam_config=/tmp/iam-config.yaml
# aws s3 cp ${remote_iam_config} ${local_iam_config}
# awsIrsaRoleArn=$(yq '.certManager.roleArn' "${local_iam_config}")
alertManagerRoleArn="arn:aws:iam::${account}:role/alertmanager-${cluster}"
o11yBackupRoleArn="arn:aws:iam::${account}:role/o11y-backup-${cluster}"
o11yBackupBucket="o11y-${env}-${logical_region_name}-backups"
nginxReadIngressClass="nginx-read"
nginxWriteIngressClass="nginx-write"

if ! aws s3api head-bucket --bucket "$o11yBackupBucket" 2>/dev/null; then
    echo "o11y bucket $o11yBackupBucket does not exist, creating..."
    aws s3 mb --region $region "s3://$o11yBackupBucket"
    echo "o11y bucket $o11yBackupBucket created"
else
    echo "o11y bucket $o11yBackupBucket already exists"
fi

args=$(cat <<EOF
{
  "alertmanagerRoleArn": "$alertManagerRoleArn",
  "o11yBackupRoleArn": "$o11yBackupRoleArn",
  "o11yBackupBucket": "$o11yBackupBucket",
  "nginxReadIngressClass": "$nginxReadIngressClass",
  "nginxWriteIngressClass": "$nginxWriteIngressClass",
  "regionalServerRoleARN": "$regionalServerRoleARN",
  "imageRegistry": "$imageRegistry",
  "hostedZoneName": "$hosted_zone_name",
  "certificateARN": "$certificateARN",
  "wafACLARN": "$wafACLARN"
}
EOF
)
echo "args: $args"


common_params="-f vendor=${vendor} -f region=${region} -f logicalRegionName=${logical_region_name} -f env=${env} -f cluster=${cluster} -f jobId=${jobId}"
# Create pull request by invoking new-region workflow.
gh workflow run new-region-infra.yaml --repo ${o11y_repo} --ref ${git_ref} ${common_params} -f args="${args}"
wait_workflow_run_to_merge new-region-infra.yaml ${o11y_repo}
if [[ $? -eq 1 ]]; then
    exit 1
fi

# Deploy application services by invoking infra-cd workflow
gh workflow run infra-cd-v2.yaml --repo ${o11y_repo} --ref ${git_ref} -f env=${env} -f region=${logical_region_name} -f vendor=${vendor} -f git-ref=${git_ref}

# Wait services ready
declare -A services

services["deploy/eventrouter -n kube-system"]="false"
services["sts/prometheus-self-monitoring-kube-prome-prometheus -n self-monitoring"]="false"
services["sts/alertmanager-self-monitoring-kube-prome-alertmanager -n self-monitoring"]="false"
services["deploy/volume-autoscaler -n self-monitoring"]="false"
services["ds/self-vector -n self-monitoring"]="false"
services["deploy/victoria-metrics-operator -n self-monitoring"]="false"

if [[ $vendor == "aws" ]]; then
  services["deploy/cert-manager -n kube-system"]="false"
fi

timeout=${JOB_TIMEOUT:=600} # 10 mins
interval=${JOB_CHECK_INTERVAL:=10}
start_time=$(date +%s)
echo "Waiting all services ready..."
set +e
while [[ $(($(date +%s) - $start_time)) -lt $timeout ]]; do
  all_ready=true
  for svc in "${!services[@]}"; do
    if [[ ${services[$svc]} == "true" ]]; then
      continue
    else
      all_ready=false
    fi

    # for self-vector daemonset, we need to check the number of available pods
    if [[ $svc == "ds/self-vector -n self-monitoring" ]]; then
      ready=$(kubectl get ${svc} -ojsonpath='{.status.numberReady}' 2>/dev/null)
    else
      ready=$(kubectl get ${svc} -ojsonpath='{.status.readyReplicas}' 2>/dev/null)
    fi

    if [[ $? -eq 0 ]]; then
      if [[ $ready -ge 1 ]]; then
        services[$svc]="true"
        echo "Service $svc is ready"
      fi
    fi
  done

  if $all_ready; then
    echo "All services ready!"
    exit 0
  else
    sleep "${interval}"
  fi
done

# Report unready services
echo "Following services are unready:"
for svc in "${!services[@]}"; do
  if [[ ${services[$svc]} == "false" ]]; then
    echo $svc
  fi
done
exit 1