Chris Christiansen
9fcb9c354a
Some checks are pending
Check Python Version Consistency / Check Python Version (push) Waiting to run
79 lines
2.6 KiB
Bash
79 lines
2.6 KiB
Bash
#!/usr/bin/env bash
|
|
# 04-observability-setup.sh — Configure Cloud Monitoring, alerting, uptime checks
|
|
# Idempotent
|
|
# Source .env before running: source .env
|
|
set -euo pipefail
|
|
|
|
: "${PROJECT_ID:?Set PROJECT_ID}"
|
|
: "${REGION:?Set REGION}"
|
|
: "${CLOUD_RUN_SERVICE:?Set CLOUD_RUN_SERVICE}"
|
|
: "${ALERT_EMAIL:?Set ALERT_EMAIL}"
|
|
|
|
echo "=== 04: Setting up observability for ${PROJECT_ID} ==="
|
|
|
|
bash "$(dirname "$0")/00-authcheck.sh"
|
|
|
|
gcloud services enable \
|
|
monitoring.googleapis.com \
|
|
logging.googleapis.com \
|
|
cloudtrace.googleapis.com \
|
|
--project="${PROJECT_ID}" --quiet
|
|
echo "✓ Monitoring APIs enabled"
|
|
|
|
# 1. Notification channel (email)
|
|
CHANNEL_FILE=$(gcloud alpha monitoring channels list \
|
|
--filter="type=email AND labels.email_address=${ALERT_EMAIL}" \
|
|
--format="value(name)" --project="${PROJECT_ID}" 2>/dev/null | head -1 || true)
|
|
|
|
if [[ -z "${CHANNEL_FILE}" ]]; then
|
|
cat > /tmp/channel.json << EOF
|
|
{
|
|
"type": "email",
|
|
"displayName": "OSVauco Alert Email",
|
|
"labels": { "email_address": "${ALERT_EMAIL}" }
|
|
}
|
|
EOF
|
|
NOTIFICATION_CHANNEL=$(gcloud alpha monitoring channels create \
|
|
--channel-content-from-file=/tmp/channel.json \
|
|
--project="${PROJECT_ID}" \
|
|
--format="value(name)")
|
|
echo "✓ Notification channel created: ${NOTIFICATION_CHANNEL}"
|
|
else
|
|
NOTIFICATION_CHANNEL="${CHANNEL_FILE}"
|
|
echo "✓ Notification channel already exists: ${NOTIFICATION_CHANNEL}"
|
|
fi
|
|
|
|
# 2. Create log-based metric for agent errors
|
|
if ! gcloud logging metrics describe agent-error-count \
|
|
--project="${PROJECT_ID}" >/dev/null 2>&1; then
|
|
gcloud logging metrics create agent-error-count \
|
|
--description="Count of ERROR severity logs from agent" \
|
|
--log-filter="resource.type=\"cloud_run_revision\" severity=ERROR" \
|
|
--project="${PROJECT_ID}"
|
|
echo "✓ Log-based metric 'agent-error-count' created"
|
|
else
|
|
echo "✓ Log-based metric already exists"
|
|
fi
|
|
|
|
# 3. Uptime check for Cloud Run service
|
|
SERVICE_URL=$(gcloud run services describe "${CLOUD_RUN_SERVICE}" \
|
|
--region="${REGION}" \
|
|
--format="value(status.url)" 2>/dev/null || echo "")
|
|
|
|
if [[ -n "${SERVICE_URL}" ]]; then
|
|
HOST=$(echo "${SERVICE_URL}" | sed 's|https://||')
|
|
gcloud monitoring uptime create \
|
|
--display-name="${CLOUD_RUN_SERVICE}-uptime" \
|
|
--protocol=HTTPS \
|
|
--request-path="/health" \
|
|
--port=443 \
|
|
--monitored-resource-type=uptime_url \
|
|
--project="${PROJECT_ID}" \
|
|
--hostname="${HOST}" 2>/dev/null || true
|
|
echo "✓ Uptime check configured for ${SERVICE_URL}"
|
|
fi
|
|
|
|
echo ""
|
|
echo "=== 04: Observability setup COMPLETE ==="
|
|
echo " Dashboard: https://console.cloud.google.com/monitoring?project=${PROJECT_ID}"
|