USB Sticks/CF Cards Should not be used as System Drive

  • Folks,


    For those who are new to Open Media Vault (OMV) please do not use a USB thumb stick or CF card for your system drive of OMV. There are many threads here in the forums about the sicks/cards failing in several days to weeks. From what I've read it is due to the way OMV writes a lot of info to the system drive. And since the flash memory in the sticks/cards is not made for the constant writing, they die in short order.


    There is an article at http://www.bit-tech.net/bits/s…/2012/07/19/replace-whs/1 that states;


    "Be warned, however: you’ll need a dedicated drive for the operating system itself. A 2GB or larger USB stick does the job nicely."


    The best part though is they praise OMV!


    "... makes OpenMediaVault a powerful alternative to Windows Home Server for media streaming."


    Good Stuff!


    Though I guess they did not run OMV for the time others have that have had problems.


    I include myself in being new to OMV as I've only been using it for the past week or so. And I did run OMV on a 16gig CF card for several of the first days without problem. I did take the advice though of those who tell us not to use sticks/cards for a system drive, and installed a barely used though old 40gig hard drive.


    There is one post here (there may be others) of someone using a high end CF card with wear leveling as part of it's construction. And they are not experiencing any issues. Which is good for them. Though wear leveling is generally not found on sticks/cards, but is on SSD drives.


    Just my three cents.


    Dave

  • My impression is that newer USB flash drives do support wear leveling, most likely dynamic wear leveling.
    I use a 2GB Toshiba USB flash drive for my OMV installation. According to Toshiba's documents, their flash memory controllers have supported wear leveling since at least 2004.
    Still, to be safe, I make a backup of OMV system before each major change.

    OpenMediaVault 7.0.4-2 (Sandworm) :thumbup:
    HARDWARE: Raspberry Pi 4 Model B Rev 1.2 | SYSTEM: Debian GNU/Linux 12 (bookworm) aarch64

    KERNEL: Linux 6.1.21-v8+ | PROCESSOR: BCM2835 (4) @ 1.500GHz | MEMORY: 4GB | STORAGE: MicroSD 60GB + HDD 6TB
    PLUGINS: nut, omvextrasorg, backup, compose, cputemp, flashmemory, kernel, remotemount, sharerootfs

    DOCKER: nginx proxy manager, pi-hole, wireguard, transmission

  • kochin


    That's good that Toshiba claims to be doing that, but is it across all their USB flash drives? Or only certain model lines? Glad that you are getting good life from yours. How long has it been in use?


    Though there are many, many posts here of USB flash drive woes. And when their flash drive fails they usually have problems, as they may not have a back up of the system drive.


    It's even better though that you have a back up of your system drive.

  • Maybe I was wrong. But increasingly I read answers on vendors' official support forums saying their USB flash drives utilize wear leveling. Some brands such as Corsair and Transcend clearly stated wear leveling as one of the features. (Strangely more recent data sheets stop mentioning wear leveling. Perhaps it has become a standard feature.)


    I am not sure whether all the models from Toshiba or other manufactures support wear leveling, but I would think they have no reason not to include it because nowadays flash controllers tend to have that feature built in.


    The USB flash drive I use for OMV is about 6 years old. However, it has only got more heavily used since last September when I tried Amahi then decided on OMV. So far I am quite happy with the setup. I did have problems with the flash drive after power outages. Now I plug my OMV box to a UPS and maintain a system backup.

    OpenMediaVault 7.0.4-2 (Sandworm) :thumbup:
    HARDWARE: Raspberry Pi 4 Model B Rev 1.2 | SYSTEM: Debian GNU/Linux 12 (bookworm) aarch64

    KERNEL: Linux 6.1.21-v8+ | PROCESSOR: BCM2835 (4) @ 1.500GHz | MEMORY: 4GB | STORAGE: MicroSD 60GB + HDD 6TB
    PLUGINS: nut, omvextrasorg, backup, compose, cputemp, flashmemory, kernel, remotemount, sharerootfs

    DOCKER: nginx proxy manager, pi-hole, wireguard, transmission

  • There are ways to reduce write operations to an USB stick or CF card.
    - mount the stick/cf as read only
    - mount a tmpfs as read write
    - bind them using unionfs or aufs (it stacks the filesystems writes will go into the tmpfs(ram) and the reads will be done from the stick/cf)


    to save the data at shutdown, you can remount the stick/cf as rw and rsync the data from the tmpfs to the stick/cf (or just copy it)


    the downside is when the power goes out the data is gone (maybe you can sync it every x minutes, so you only loose x mins of data)


    some more info can be found here:
    http://www.logicsupply.com/blo…a-read-only-linux-system/
    https://bbs.archlinux.org/viewtopic.php?id=133823
    http://jrfom.com/2012/05/09/pr…ife-by-mitigating-writes/

  • Zitat von "StrikeByte"

    There are ways to reduce write operations to an USB stick or CF card.
    - mount the stick/cf as read only


    Strike, I read stuff like this and want to delete your post. I will just say that OMV cannot operate like this....

  • Is this really worth the added risk and time? We recommend a hard drive, or SSD, for a reason. A lot of people are not even using a battery backup. We have had many issues with flash as system drives and this adds a great deal of risk/complexity to a system. We do not recommend this type of a setup.

  • And the main reason for now is, that OMV is not built for using it of of FLASH USB keys. This may change, but for now it is not.

    Everything is possible, sometimes it requires Google to find out how.

  • Seagate article confirms low write capabilities. Please read http://www.seagate.com/point-o…supply-market-master-pov/

    • Offizieller Beitrag
    Zitat von "manu512"

    What about using F2FS as filesystem for the system in the future ?


    F2FS first appears in the 3.6 kernel but really you want 3.8 or later (maybe even 3.9 or 3.10). Even Wheezy's kernel is only 3.2 and I'm not sure if F2FS is/will be backported to it.

    omv 7.0.5-1 sandworm | 64 bit | 6.8 proxmox kernel

    plugins :: omvextrasorg 7.0 | kvm 7.0.13 | compose 7.1.4 | k8s 7.1.0-3 | cputemp 7.0.1 | mergerfs 7.0.4


    omv-extras.org plugins source code and issue tracker - github - changelogs


    Please try ctrl-shift-R and read this before posting a question.

    Please put your OMV system details in your signature.
    Please don't PM for support... Too many PMs!

  • Don't know that I would still trust it for OMV or other write intensive applications. And now that I've had time to think about the article, don't know that I would want to use flash memory for more than possibly storage. Why?


    Based on a 3000 write cycle lifetime (from the article) that does not seem like a long time. Particularly when the flash drive may be nearly full. Which means more writes to the same memory addresses. Granted there are controllers that look at how many writes have occurred per memory address and will move data from low usage areas but how do we as consumers know that our $8.00 8gig flash stick has that capability? I suspect the high end controllers cost more and end up in the enterprise/business class of SSD's. Yes some may end up in high end USB/CF devices again though how do we know.


    A quick check shows refurb Sata drives for about $30.00 new ones for about $55.00 Yes more than a flash drive. But what is your time and data worth?


    As with anything man made - It's not a matter of if it will fail. But when it will fail. So back up, back up, and back up once more!

    • Offizieller Beitrag

    I'm not 100% sold on SSD's yet. First, there's reliability. Some seem to work really well, others don't. Unfortunately my only experience with them, was one that didn't. Second, I think the cost per GB is still to high compared to platter drives.


    Frankly, there's an abundant number of places to get "old", new in box drives that are more than capable of hosting the OS. Just a quick glance at Ebay, and I see several new 2.5in hard drives by Seagate, Hitachi and WD... they are in the $25-$40 range, and range in size from 40-80gigs. Yeah, my NAS may boot 10-15sec slower because I went this route, but when something works, I don't mess with it.

  • It is not about slow. If you want a non spinning low power disk, a SSD is what you can use. A USB stick would be even better, but that is not recommended. Of cause a physical disk is cheaper, but that was not the point here.

    Everything is possible, sometimes it requires Google to find out how.

  • To be "green" enough, my hardware need only 18W of power supply in no load condition. Because I need to use an USB stick for the system - datas goes on the 2 only possible drives mirrored -, I was looking for a usable solution. This is explained here: http://bugtracker.openmediavault.org/view.php?id=526. The script given by votdev, modified for EXT4 works well. Only left the problem with the start of the ramlog demon.
    After hours of study, I found a solution, given in this post: http://bugtracker.openmediavault.org/view.php?id=526#c2104
    A copy of logs is made each hours, and the USB key is no more burned by too frequent writes. My SAN worked without problems for 4 month.

Jetzt mitmachen!

Sie haben noch kein Benutzerkonto auf unserer Seite? Registrieren Sie sich kostenlos und nehmen Sie an unserer Community teil!