Proxmox SSD Disappearing After Reboot: Troubleshooting Guide

You know it's a bad sign when your server ghosted you midweek and hasn't shown signs of life since. Worse yet? You're thousands of miles away from your rig, and the usual fix—a reboot—falls flat. Welcome to the latest anxiety-inducing chapter of homelab horror stories: the case of the disappearing SSD in Proxmox. This isn't a theoretical problem. A user recently shared their real-life tech troubles involving a Dell T5810 server running Proxmox with a PCIe NVMe card—three out of four slots populated, with one crucial drive mysteriously dropping out. It's the one holding the VMs and LXC containers. The heart of the operation. And it's just… gone. Let's break down what happened, why this might be happening, and what the community had to say—because this issue, it turns out, is a lot more common than it should be. ## The Setup: All Good Until It Isn't Everything started off clean. Fresh server build, Proxmox humming, and three NVMe drives in action. But then, the main storage SSD started randomly vanishing. Once? Maybe just a fluke. Twice? You start raising eyebrows. And now, after being offline for several days, not even a standard restart brings it back. The only "solution" so far? A full shutdown—pull the plug, reinsert power—and poof, it's back. Until it happens again. That's where things get murky. Is it a a software issue? A failing SSD? A power delivery problem? Or something sneakier happening with PCIe lanes and oversaturation? ## Step One: Check the Obvious (and Not-So-Obvious) ### 1. Bad Drive? Maybe Multiple folks suggested the first line of defense: check if the SSD itself is just dying. Run smartctl diagnostics and look for wear-level indicators. You'd be surprised how quickly commercial SSDs can burn out under heavy virtualization workloads. Proxmox, especially if using ZFS, is notorious for shredding consumer-grade flash storage. But our poster clarified—they're using a Fanxiang 1TB NVMe, and while it's not enterprise-grade, it's not totally off-brand either. One commenter claimed solid results with multiple Fanxiang S880s, suggesting the drive itself might not be the villain here. Still, wear happens. And if you're seeing 90% wear on your drive after just a year or two, that's a red flag waving at full mast. ### 2. ASPM Settings & PCIe Quirks Another user suggested toggling ASPM (Active State Power Management) in BIOS. Turns out, ASPM can cause connectivity issues with certain SSDs when power management kicks in. It sounds like low-level tinkering, but if you're running a PCIe expansion card and only one drive keeps dropping, this might just fix it. And yes, ASPM can affect just one of the SSDs on a multi-slot card—this isn't an all-or-nothing setting. Different drives, even on the same card, might react differently to BIOS-level power controls. ### 3. Oversaturated PCIe Bus One user had a similar disappearing act that ended when they yanked out one of their four add-in cards. Too many PCIe devices—GPUs, network cards, NVMe arrays—can crowd your lanes and starve some devices, especially if you're on a platform not designed for heavy-duty expansion. Running two GPUs and a NIC along with an NVMe adapter? That might be pushing it. Drop a GPU and see what happens. If the ghosting drive suddenly reappears and stays solid, you've just solved a bandwidth problem disguised as hardware failure. ## Other Suspects in the Lineup ### Loose PCIe Card It's not always high-tech. One person found that their PCIe card just wasn't seated properly. Vibration, travel, or thermal expansion could wiggle it loose over time. Reseating the card and SSDs—fully unplugging and replugging—might be boring, but it often works. Add in some proper cable management or a bracket to keep the card steady, especially in vertically-mounted configurations. ### Bad Cooling Yes, seriously. Overheating SSDs or cards without proper airflow can throttle, disconnect, or outright vanish under load. One user fixed repeated dropouts by aiming a simple fan directly at their NVMe card. That's it. A $10 USB fan doing what BIOS updates and firmware patches couldn't. ### PSU and Power Rail Issues Here's a curveball: power delivery. If you're splitting power cables between GPUs and SSDs, you might be overloading a rail or causing voltage instability. One guy fixed his issue by simply running separate power cables to each 8-pin GPU connector—something that should've been obvious in hindsight but made all the difference. ## The Fixes That Actually Worked So what can you do when you're stuck in this Proxmox limbo? - **Update SSD firmware**: If your model has known bugs (Samsung 990 Pro, we're looking at you), a firmware patch might fix random disconnects. - **Check the card seating**: Reseat the PCIe NVMe adapter and each drive. - **Tweak BIOS settings**: Disable ASPM or try changing PCIe slot configurations. - **Drop extra cards**: Free up PCIe bandwidth by removing unused or secondary expansion cards. - **Cool it down**: Add a fan directly blowing on the PCIe NVMe adapter. - **Swap the power cables**: Use separate GPU cables instead of daisy-chaining. - **Mount logs to RAM**: If Proxmox is writing constantly to SSDs, move /var/log to tmpfs to reduce write cycles. - **Check drive health**: Use smartctl and other tools to monitor SSD wear and performance stats. ## Final Thoughts: Homelab Wisdom from the Trenches This disappearing SSD issue is part hardware, part power, part software—classic homelab chaos, basically. The more complex your setup, the more points of failure sneak in. But the community, as usual, pulls through with a grab bag of troubleshooting tips that range from the obvious to the obscure. The big takeaway? Always assume it's the simplest thing first. A loose card, a heat problem, or a power delivery quirk might save you hours of debugging logs and diving into kernel messages. Also: keep spares, monitor your drives regularly, and if you're running anything mission critical on non-enterprise SSDs... maybe rethink that. Or at least keep a backup plan that doesn't involve a 12-hour flight home just to reseat a card.