easyai-ai-gateway/scripts/cluster/verify-cluster.sh

#!/usr/bin/env bash
set -euo pipefail

script_dir=$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)
# shellcheck source=scripts/cluster/common.sh
source "$script_dir/common.sh"

mode=${1:-precutover}
[[ $mode == precutover || $mode == postcutover ]] || {
  echo 'usage: verify-cluster.sh [precutover|postcutover]' >&2
  exit 64
}
load_cluster_env

hosts=("$CLUSTER_NINGBO_HOST" "$CLUSTER_HONGKONG_HOST" "$CLUSTER_LOS_ANGELES_HOST")
wireguard_ips=(10.77.0.1 10.77.0.2 10.77.0.3)
for source_index in 0 1 2; do
  handshakes=$(cluster_ssh "${hosts[$source_index]}" \
    "wg show wg0 latest-handshakes | awk '\$2 > 0 && systime() - \$2 < 180 { count++ } END { print count + 0 }'")
  [[ $handshakes -eq 2 ]] || {
    echo "stale WireGuard handshake on ${hosts[$source_index]}" >&2
    exit 1
  }
  for target_index in 0 1 2; do
    [[ $source_index -eq $target_index ]] && continue
    ping_output=$(cluster_ssh "${hosts[$source_index]}" \
      "ping -c 10 -W 2 ${wireguard_ips[$target_index]}")
    packet_loss=$(sed -nE 's/.* ([0-9.]+)% packet loss.*/\1/p' <<<"$ping_output")
    average_rtt=$(sed -nE 's#.* = [0-9.]+/([0-9.]+)/.*#\1#p' <<<"$ping_output")
    awk -v loss="$packet_loss" 'BEGIN { exit !(loss < 1) }'
    if [[ $source_index -eq 2 || $target_index -eq 2 ]]; then
      awk -v rtt="$average_rtt" 'BEGIN { exit !(rtt < 300) }'
    else
      awk -v rtt="$average_rtt" 'BEGIN { exit !(rtt < 80) }'
    fi
  done
done
echo 'wireguard_acceptance=PASS handshakes=6 packet_loss_lt_1=true'

cluster_state=$(cluster_ssh "$CLUSTER_NINGBO_HOST" 'bash -s' <<'REMOTE'
set -euo pipefail
kubectl='k3s kubectl'
[[ $($kubectl get nodes --no-headers | awk '$2 == "Ready" { count++ } END { print count + 0 }') -eq 3 ]]
$kubectl get --raw='/readyz?verbose' | grep -q '\[+\]etcd ok'
[[ $($kubectl get node easyai-los-angeles -o json |
  jq '[.spec.taints[]? | select(.key=="easyai.io/witness" and .effect=="NoSchedule")] | length') -eq 1 ]]
[[ $($kubectl get pods -A -o json |
  jq '[.items[] | select(.spec.nodeName=="easyai-los-angeles") |
    select(.metadata.namespace != "kube-system" and .metadata.namespace != "cnpg-system" and .metadata.namespace != "cert-manager")] | length') -eq 0 ]]
printf 'nodes=%s\n' "$($kubectl get nodes --no-headers | wc -l)"
REMOTE
)
[[ $cluster_state == nodes=3 ]]
echo 'k3s_control_plane=PASS servers=3 witness_business_pods=0'

database_state=$(cluster_ssh "$CLUSTER_NINGBO_HOST" 'bash -s' <<'REMOTE'
set -euo pipefail
kubectl='k3s kubectl'
[[ $($kubectl get cluster easyai-postgres -n easyai -o jsonpath='{.status.readyInstances}') -eq 2 ]]
primary=$($kubectl get cluster easyai-postgres -n easyai -o jsonpath='{.status.currentPrimary}')
[[ $($kubectl get pod "$primary" -n easyai -o jsonpath='{.spec.nodeName}') == easyai-ningbo ||
   $($kubectl get pod "$primary" -n easyai -o jsonpath='{.spec.nodeName}') == easyai-hongkong ]]
replication=$($kubectl exec -n easyai "$primary" -c postgres -- \
  psql -X -U postgres -d easyai_ai_gateway -At -c \
  "SELECT count(*) FROM pg_stat_replication WHERE sync_state IN ('sync','quorum');")
[[ $replication -eq 1 ]]
archived_count=$($kubectl exec -n easyai "$primary" -c postgres -- \
  psql -X -U postgres -d postgres -At -c \
  'SELECT archived_count FROM pg_stat_archiver;')
(( archived_count > 0 ))
[[ $($kubectl get cluster easyai-postgres -n easyai \
  -o jsonpath='{.status.conditions[?(@.type=="ContinuousArchiving")].status}') == True ]]
completed_backups=$($kubectl get backups -n easyai -o json |
  jq '[.items[] | select(
    .spec.cluster.name == "easyai-postgres" and .status.phase == "completed"
  )] | length')
(( completed_backups > 0 ))
for pod in $($kubectl get pods -n easyai -l cnpg.io/cluster=easyai-postgres \
  -o jsonpath='{.items[*].metadata.name}'); do
  $kubectl exec -n easyai "$pod" -c postgres -- \
    test -r /projected/barman-aws/config
done
printf 'primary=%s archived=%s backups=%s\n' \
  "$primary" "$archived_count" "$completed_backups"
REMOTE
)
[[ $database_state == primary=easyai-postgres-* ]]
echo "postgres_preflight=PASS $database_state synchronous_replicas=1"

if [[ $mode == postcutover ]]; then
  application_state=$(cluster_ssh "$CLUSTER_NINGBO_HOST" 'bash -s' <<'REMOTE'
set -euo pipefail
kubectl='k3s kubectl'
[[ $($kubectl get pods -n easyai -l app.kubernetes.io/name=easyai-api \
  --field-selector=status.phase=Running --no-headers | wc -l) -eq 2 ]]
[[ $($kubectl get pods -n easyai -l app.kubernetes.io/name=easyai-web \
  --field-selector=status.phase=Running --no-headers | wc -l) -eq 2 ]]
images=$($kubectl get deployments -n easyai -o json |
  jq -r '.items[] | select(.metadata.name | startswith("easyai-")) |
    .spec.template.spec.containers[].image')
if grep -Ev '@sha256:[0-9a-f]{64}$' <<<"$images" | grep -q .; then
  echo 'one or more application images are not digest-pinned' >&2
  exit 1
fi
[[ $($kubectl get secret easyai-gateway-security-events -n easyai \
  -o json | jq '.data | length') -eq 4 ]]
[[ $($kubectl auth can-i update secret/easyai-gateway-security-events -n easyai \
  --as=system:serviceaccount:easyai:easyai-ai-gateway) == yes ]]
[[ $($kubectl auth can-i update secret/easyai-ai-gateway-runtime -n easyai \
  --as=system:serviceaccount:easyai:easyai-ai-gateway) == no ]]
printf 'applications=ready\n'
REMOTE
)
  [[ $application_state == applications=ready ]]
  echo 'application_acceptance=PASS api=2 web=2 digests=pinned'

  for edge_ip in "${CLUSTER_NINGBO_HOST#*@}" "${CLUSTER_HONGKONG_HOST#*@}"; do
    curl -fsS --max-time 10 --resolve "ai.51easyai.com:443:$edge_ip" \
      https://ai.51easyai.com/api/v1/healthz | grep -q easyai-ai-gateway
    curl -fsS --max-time 10 --resolve "ai.51easyai.com:443:$edge_ip" \
      https://ai.51easyai.com/api/v1/readyz | grep -q '"ok":true'
    curl -fsS --max-time 10 --resolve "ai.51easyai.com:443:$edge_ip" \
      https://ai.51easyai.com/api/v1/openapi.json | grep -q '"/api/v1/chat/completions"'
    curl -fsS --max-time 10 --resolve "ai.51easyai.com:443:$edge_ip" \
      https://ai.51easyai.com/ | grep -q 'EasyAI AI Gateway'
  done
  echo 'nginx_edges=PASS ningbo=true hongkong=true dns_changed=false'
fi

echo "cluster_verification=PASS mode=$mode"