Lolight

This DOM has been recently purchased from DigiKey: https://www.digikey.com/en/products/detail/atp-electronics-inc/AF16GSMGH-AABXP/16374995 ATP DOM - 16 GB, eUSB P/N: AF16GSMGH-AABXP Year: 2025 GUID: UNKNOWN (was not tested) Controller: Silicon Motion SM3257 ENAA Possible Memory Chip(s): Toshiba TC58TFG7DDLTA00, 15nm Memory Type: MLC Flash ID: 983A9493 7651 Flash CE: 1 Chip F/W: ISP 160729-AA0 MP: P0729BT MPTOOL Ver.: 2.05.51 PTEST Date: 2014-12-19 VID: 14AF PID: AF01 Manufacturer: ATP Electronics Product: ATP CG eUSB Physical Disk Capacity: 16 GB Relative Offset: 56 KB USB Version: 2.00 Declared Power: 500 mA

My understanding is that it's pretty much the same thing as with SSDs. Both methods can have a partition allocated for storage it's just not as practical for USBs drives.

Each device in internal boot consumes a license slot. With mirrored USB flash drive in internal boot config they will consume 2 license slots. The license key can be placed on either TPM or third USB drive - your choice. The third USB drive with the license key on it won't consume a license slot.

Mirrored USB drives in a pool is one of the available internal boot configurations. You can't mirror USB flash drives while on USB boot.

One thing worth remembering is that internal boot has several drawbacks. Read up on it first -- the USB boot guide provides lots of relevant info on USB flash drives and internal boot configurations. Then decide for yourself. Besides, it's still in beta - wait for stable.

One thing worth remembering is that internal boot has several drawbacks and only a few advantages. Read up on it first, then decide for yourself. Marketing materials don’t tell the whole story. A good place to start is the USB Flash guide in the Boot Devices category of this forum.

Same boat here -- ~ 50TB offsite backup server is exactly what I'm trying to figure out myself. Don't want to run any complicated custom setups. For the power on/off question ConnerVT's thread linked above covers it well -- WOL to wake the backup server, S3 sleep plugin to send it back to sleep when done. Works reliably without needing IPMI hardware. For the sync software I've been looking at three options and haven't committed yet. Syncthing is simple -- zero configuration networking, bidirectional sync, web UI only, no CLI required. Resilio Sync is very user friendly but proprietary. Duplicacy has better versioning and chunk-based deduplication which protects against the kind of ransomware scenario where a simple mirror would just replicate the encrypted files. One thing worth knowing if Duplicati (not Duplicacy) comes up as a recommendation — Spaceinvaderone's tutorial on it has a disclaimer added recently noting restore reliability issues. Worth reading before going that route. Duplicacy is the more reliable alternative in the same general space. Still weighing the options myself. Would be interested to hear what others in a similar situation ended up choosing -- especially at the >50TB scale where the initial sync alone is a multi-day project regardless of what software you pick.

The drive is clean and the symptoms were most likely a temporary system state issue. That said, the drive has been in service since July 2017 -- nearly 9 years. At 32GB it's almost certainly planar TLC which is adequate but not the most reliable option long term. Worth running ChipGenius to confirm what's inside. Depending on what you find, sourcing a verified quality legacy MLC drive as a replacement might be the simpler and more reliable path compared to migrating to internal boot -- the USB Flash section guide and MLC companion thread cover exactly what to look for and list specific verified options currently available on eBay. On internal boot -- beta.2 just dropped and the bug backlog is still being worked through. Your pool drives will still be there when stable releases. No need to rush the migration.

The invoice download working confirms your purchase is on record. The order confirmation link error and the missing key file in your account are something support will need to resolve directly -- the account UI clearly isn't surfacing the license correctly for your order. Contact Unraid support at [email protected] with your purchase date, the email address used for the purchase, and the downloaded invoice PDF attached. They can locate the license key manually and send it to you. Once you receive it copy the .key file into the /boot/config folder on your boot drive and the registration page should recognize it immediately.

The process starts with backing up your existing drive through Main --> Boot Device --> Boot Device Backup. Then use the USB Flash Creator -- select "Use custom" option and point it at the backup ZIP file. This copies your existing configuration and license key across to the new drive rather than writing a fresh installation. Boot from the new drive. Go to Tools --> Registration and click Replace Key. This transfers the license association from the old drive's GUID to the new drive's GUID on Lime Technology's license server. The old drive gets blacklisted automatically. The automated transfer is available once per year. If you need another within 12 months contact support directly. Worth checking the USB Flash section guide before selecting your new drive -- it covers what makes a reliable boot drive for always-on server duty.

Samsung's consumer USB drives are actually one of the more commonly recommended but not particularly suitable options for always-on Unraid boot duty -- so the failures you experienced aren't random bad luck or the nature of USB flash generally. The Bar Plus and FIT Plus contain modern NAND that requires continuous idle controller maintenance regardless of how they're mounted or how rarely you reboot. The UPS protects against corruption from power loss but doesn't address the underlying NAND stability issue. The drives that run for decades in Unraid are overwhelmingly older large-node MLC hardware from the 2007-2012 era -- fundamentally more stable NAND that requires essentially zero idle controller activity. The USB Flash section guide in this forum covers exactly why and what to look for. Glad internal boot is working well for you. Worth checking what NAND is in your internal drive -- if it's a modern consumer NVMe the same physics apply in a different form factor. The guide covers that too.

There's no mystery behind why some drives offer long-term reliability while others suffer from disastrous performance. You can easily assess your Verbatim drive's potential longevity by checking its components -- it takes just a few minutes.

Just to clarify -- in your original post you only mentioned rearranging drives and cables, with no mention of introducing any new cables into the modular ports of the PSU. Even if that spare cable is genuinely a Corsair cable, Corsair modular cables are frequently not interchangeable between their own models. The CX, RM, HX and other series have used different pinouts on physically identical connectors at various points, and even different generations of the same model can differ.

To expand on JorgeB's comment above: With USB licensing: Drive 1 -- USB flash drive or DOM, 8GB minimum, boot pool member Drive 2 -- USB flash drive or DOM, 8GB minimum, boot pool member Drive 3 -- existing USB license drive, with no size restrictions, holds the license key as before With TPM licensing: Drive 1 -- USB flash drive or DOM, 8GB minimum, boot pool member Drive 2 -- USB flash drive or DOM, 8GB minimum, boot pool member License -- stored on dTPM or fTPM. As a suggestion, you might want to consider these dirt cheap yet very solid low end industrial drives which are likely to contain 15nm MLC -- eBay item number: 326046070546. If you get them, please run ChipGenius and share the output.

Yes, Tools --> New Config is the correct starting point to reassign your remaining drives and rebuild your array configuration. One thing worth addressing before you do -- that generic USB drive is a liability. Back it up now through Main --> Boot Device --> Boot Device Backup. Then start looking for a replacement while the system is stable -- not when it fails. The USB Flash section guide covers what to look for and the companion thread lists specific verified MLC drives on eBay with item numbers ready to go.

Possibly a failing PSU. If it is then the most important immediate step is to stop using that PSU entirely to avoid killing the remaining drives. Beyond that, professional data recovery is likely the only realistic path since electrically damaged drive PCBs are generally beyond DIY repair unless you can source an identical PCB donor with a matching firmware chip — and even that is hit or miss. The CX450M has had a troubled history on PSU tier lists. Actually this particular model is quite decent overall, rated at C+ on the PSU tier list which is quite OK for a value priced PSU. The only thing is that according to this review it has excessively high OCP on the minor rails https://www.techpowerup.com/review/corsair-cx-m-series-450-w/12.html It is a critical safety compromise. From the review: "OCP on the minor rails is set very high, especially at 3.3V. This is not good at all, as it can lead to damaged components in case something goes wrong."

Those old drives are not HAMR.

These Barracudas carry a laser warning indicating HAMR technology -- the same Mozaic platform Seagate uses for Exos. The physical hardware points to the same underlying drives: a 30TB 10-platter Exos with one or two defective head pairs gets firmwared down to 24TB or 22TB; a 28TB 8-platter Exos that fails speed specs but has intact heads ships as a 28TB Barracuda. Different failure modes, different salvageable capacities -- all landing under the Barracuda label. It's standard binning practice, just applied to a full drive platform rather than a chip. You're right that Barracuda isn't rated for NAS -- but that rating reflects warranty tier and workload/speed qualification, not fundamentally different hardware. Seagate hasn't clarified any of this publicly, which is telling in itself. Whether you'd trust one in a NAS is a personal risk call, but "completely different drives" isn't accurate at these capacities.

Don't know all details. The official Unraid documentation specifically states "Ensure only one key file exists in the config folder". https://docs.unraid.net/unraid-os/system-administration/maintain-and-update/changing-the-flash-device/

The EGUID error is a key file placement issue. Your syslog confirms you purchased an Unleashed license -- so the file should be named Unleashed.key inside the /boot/config folder on the drive. With the drive accessible from another computer navigate to that folder and verify the file is present and correctly named. If it's missing, download it directly from account.unraid.net and place it there manually. Support will tell you the same thing -- straightforward fix once the file is confirmed present. HWiNFO won't add anything beyond what ChipGenius and Flash Drive Information Extractor have already provided -- it only shows VID, PID and serial number with no visibility into the controller or NAND. Nothing useful there. On H2testw -- Windows only, and only worth running if you're planning to redo your Unraid installation afterwards. The test fills the entire drive with data which will overwrite your OS files, configuration and license key completely. Since your drive is now set up and working there's no practical way to run the speed test without destroying what's on it.

Appreciate the share! Seeing a Swissbit SLC here is a treat. The price is obviously overkill for a desk-side server, but for anyone running Unraid at a remote site this is the play. When a "truck roll" to go swap a failed consumer drive costs more than the hardware, paying for industrial SLC is just smart business. It’s the ultimate buy once, cry once boot drive.

An excellent find! Did you buy it last Friday? 🙂 I think it was the only 4GB Optima Pro that remained unsold on eBay for the last month or so. Both drives utilize the Phison PS2231 controller and identical firmware, but they differ in their Toshiba MLC NAND nodes. USA Model (91005813F-4GB-U) -- 70nm NAND (TH58NVG5D4CTG20) with 2KB pages. This node offers a theoretical advantage in random/small writes and superior long-term data retention due to larger physical cell size. China Model (91005813F-4GB-C) -- 56nm NAND (TC58NVG5D1DTG20) with 4KB pages. This node is theoretically more efficient for sequential writes, as it processes twice the data per program cycle. In an Unraid boot environment, performance differences are expected to be negligible, likely within a 1–3 MB/s range. The drives report different IDs (154B:0015 vs. 0930:6545) due to separate manufacturing and supply chain modes. VID 154B (PNY): This drive was assembled at PNY’s New Jersey facility. The factory received populated PCBs likely of the same exact origin as in the Chinese drive and performed final assembly, including soldering the USB connector. Consequently, the controller was programmed with PNY’s native Vendor ID. VID 0930 (Toshiba): This is a white-labeled (rebranded) Toshiba OEM build. PNY likely contracted a Chinese facility already mass-producing Toshiba-branded drives -- VID 0930 is exclusive to Toshiba. PID 6545 identifies the Toshiba TransMemory series. To maintain production efficiency, PNY utilized the existing at that assembly line Toshiba firmware configuration rather than re-flashing to PNY identifiers. In short, the different VID and PID reflect supply-chain logistics, not meaningful differences in component quality, core performance or long-term reliability for Unraid use. Both are excellent boot devices. If you have a chance sometime, would you mind running a quick H2testw write/read test on your drive and sharing the speeds? I’d love to see how the 70nm version performs compared to the 56nm one. Also would you mind if I posted your drive’s full component output in the existing thread right under my China drive output (for easy comparison)? I can edit my original post to keep everything neatly together. Let me know if that’s okay with you.

Don't... I'd advise to check the guide first.

With the dedicated boot pool option in beta.2 eliminating the forced data partition — does the 16GB minimum size requirement still apply to USB drives and DOMs used as dedicated boot pool members, or can smaller drives now be used?