We are experiencing intermittent false “website down” alerts in Zabbix while monitoring websites that are proxied through Cloudflare. The behavior suggests a potential interaction issue between Cloudflare proxying and Zabbix web monitoring, rather than actual site downtime.
1. Environment
2. Monitoring Configuration
Zabbix web scenario settings:
Monitoring is performed using Zabbix web scenarios, not ICMP or agent-based checks.
3. Observed Behavior
4. Behavior When Cloudflare Proxy Is Disabled
5. Server Health During Alerts
During alert windows, we observed:
6. Prior Cloudflare Feedback
Cloudflare has confirmed:
7. NGINX Configuration Notes
8. Key Observation
Despite alerts:
9. Assistance Requested
We would appreciate guidance on:
1. Environment
- Hosting: DigitalOcean
- Web Server: NGINX
- OS: Ubuntu
- Application: WordPress
- CDN / Proxy: Cloudflare (orange cloud enabled)
- Monitoring: Zabbix Web Monitoring (HTTP/HTTPS scenarios)
2. Monitoring Configuration
Zabbix web scenario settings:
- Update interval: 90 seconds
- Attempts: 2
- Timeout: 45 seconds
Monitoring is performed using Zabbix web scenarios, not ICMP or agent-based checks.
3. Observed Behavior
- Zabbix triggers intermittent “website down” alerts
- Alerts occur only when Cloudflare proxy is enabled
- Alerts do not trigger immediately after enabling the proxy
- Alerts appear randomly (sometimes persist for hours)
- Alerts often self-resolve within minutes
- Real users can access the websites normally during alert periods
4. Behavior When Cloudflare Proxy Is Disabled
- Alerts stop initially
- In some cases, alerts still appear later but far less frequently
5. Server Health During Alerts
During alert windows, we observed:
- No high CPU, memory, disk, or network usage
- No spikes in NGINX error logs
- No PHP-FPM crashes or restarts
- No firewall rules blocking HTTP/HTTPS traffic
- Fail2ban is enabled only for SSH
6. Prior Cloudflare Feedback
Cloudflare has confirmed:
- No incidents or maintenance during alert times
- No visible errors from the Cloudflare side
- Enabling/disabling the proxy has a direct impact on alert frequency
7. NGINX Configuration Notes
- Previously used limit_req_status 444
- Changed to 429
- Alerts continue even after removing 444 responses
- No rules explicitly blocking Cloudflare IP ranges
8. Key Observation
Despite alerts:
- Websites remain reachable
- Alerts appear to be false positives
- Behavior strongly suggests a monitoring/proxy interaction issue, not real downtime
9. Assistance Requested
We would appreciate guidance on:
- Known issues between Cloudflare proxy and Zabbix web scenarios
- Potential impact of:
- IPv6
- TLS / SNI handling
- Expected HTTP status codes
- Cloudflare edge behavior vs monitoring probes
Comment