I stand corrected on the procedure, but maintain the question: Is it worth the hassle for a limited amount of non-redundant storage (never to be used for anything important)?
If you have loads of SATA ports or PCIe slots to add more ports, don’t do this.
If you have even one SATA port spare don’t do this.
If you are a commercial organisation or the server is mission critical or any sort of down time is a real problem, don’t do this.
But if you genuinely are out of SATA ports and cannot have separate boot and apps SSDs (and cannot afford to spend money on replacement hardware), and if it is a home environment where down time is a minor inconvenience, then IMO this is not a bad solution.
1 Like
If you’re a home user genuinely out of SATA ports
AND out of M.2 slots, PCIe slots, including x1 slots,
AND out of USB ports to deport a SAT/NVme boot drive on adapter…
I used to use Samsung Fit USB drives for ESXi and FreeNAS.
They were embedded systems so the writes to the drive were minimal and they worked well and lasted long enough (couple years for FreeNAS, quite longer for ESXi).
Things changed with SCALE as they do use the boot drive heavily at times, and, if you want full (local) Logs and other niceties, you need a ‘real’ drive.
Don’t waste money in the smaller SSD-like USB drives.
I also have an SSK USB drive that works well and it’s fast, but it is expensive for what you get.
Just buy a $10 2.5in USB enclosure (most should be ok, but watch for the cheaper ones) and a $20 decent SSD. Or an NVMe enclosure and SSD.
Most servers are old enough (before booting from NVMe) so adding a PCIe adapter with an NVMe for Boot, is not a solution (I do have a couple of them, but used as data/vm pools).
I’m using older 2.5in Samsung SSD drives in most of my and customer servers (10 out of 12?) and have not had disconnects or any issues while running.
The only possible issue I found, is right after a TrueNAS upgrade:
On 2 machines, the boot drive is not found after the restart from the upgrade.
This is either a BIOS setup or bug issue, or related to the specific USB enclosure (cheaper kind without the extra capacitor, I am thinking).
Such a small issue that I have not needed to dig deeper, as a simple control-alt-delete (or power off/on) makes the machine boot again without issues.
Few machines have ILO so I can restart them remotely, or I can simply plan the upgrade while I am physically around.
Yes, the upgrades can take a bit longer, as the USB layer adds some latency for small writes, but for the convenience and the fact that you can have an extra SATA port for your pools, makes it a good solution for me, and I will continue using them.
*One other possible issue could be if you have Virtual Machines and want to save VM States before upgrades, reboots, shutdowns or power outages.
The “default” is to save the VM state to a path (/var/lib/libvirt/qemu/save) in the Boot drive, which could create issues if 3 or 4 VMs are being saved at once.
I solved that issue with an Init script that creates a soft link to a directory in a pool in an NVMe drive, but I still WISH that TrueNAs would have a setting to SUSPEND VMs instead of the hard default of SHUTDOWN!!!