Hardware: Dell OptiPlex 7090, i5-11500T, 32GB RAM, 1TB NVMe, Proxmox VE 9.1.3, kernel 6.17.13-13-pve
Problem: Every few days the host becomes completely unresponsive — Proxmox web UI dead, SSH dead, Glances dead — but the host responds to ping normally. Hard power cycle is the only recovery. Has happened 6+ times over several weeks.
What the logs show: Nothing. Journal cuts off cleanly with zero errors at the moment of failure. No OOM events, no kernel panic, no MCE errors, no hardware errors. Persistent journald logging enabled, same result.
What I've ruled out:
Question: Has anyone seen this "pingable but dead" pattern on Proxmox? What could keep the kernel TCP stack alive while killing all userspace services with zero log entries?
Problem: Every few days the host becomes completely unresponsive — Proxmox web UI dead, SSH dead, Glances dead — but the host responds to ping normally. Hard power cycle is the only recovery. Has happened 6+ times over several weeks.
What the logs show: Nothing. Journal cuts off cleanly with zero errors at the moment of failure. No OOM events, no kernel panic, no MCE errors, no hardware errors. Persistent journald logging enabled, same result.
What I've ruled out:
- OOM / memory pressure (31GB available, no swap used)
- Kernel 7.x OpenVINO GPU driver issue (downgraded to 6.17, still crashes)
- UAS USB driver issue (fixed, still crashes)
- Firewall blocking connections (disabled)
- Hardware watchdog (no /dev/watchdog device)
- IPMI hardware event log (OptiPlex has no IPMI)
- NVMe corruption (smartctl shows 0 media errors)
Question: Has anyone seen this "pingable but dead" pattern on Proxmox? What could keep the kernel TCP stack alive while killing all userspace services with zero log entries?
