web1 502 Bad Gateway — PHP-FPM becomes unresponsive requiring container restart

[kbcmdba]

Summary

PJS2 became inaccessible (502 Bad Gateway from nginx) on 2026-04-23. The PHP-FPM backend stopped responding while nginx remained up. sudo reboot from within the container also failed (Connection timed out from systemd/dbus). Required a pct stop 128 && pct start 128 from the Proxmox host (pm1) to recover.

Impact

• PJS2 web UI and REST API both unavailable
• No data loss observed after restart
• Unknown root cause — may be PHP-FPM process exhaustion, memory pressure, or systemd hang

Proxmox warning on restart

WARN: Systemd 252 detected. You may need to enable nesting.

This nesting warning may be related to the hang.

Suggested investigation

• Check PHP-FPM pool config for max_children / process limits
• Check container memory limits vs actual usage before crash
• Consider enabling nesting for container 128 (pct set 128 --features nesting=1)
• Add Zabbix monitoring for PHP-FPM process count and response time
• Consider adding a health check endpoint (e.g., /api/health.php) that Zabbix can poll