Rock64 rant

  • Hi,
    Sorry if this isn't new, though I doubt it...
    (my problem is pretty weird + I haven't figured out where the search button is on this forum yet so you'll have to excuse me...)


    So I bought the rock64 2GB, "v2" apparently, from Amazon hoping to run OMV and get a great cheap little NAS...
    received it about a week ago


    Apart from the rock64 with a plastic case and power supply,
    I used 2 external usb 3 drives - transcend storejet 2TB & a WD Elements 1TB + usb 3 hub (also from Transcend)to connect them.
    and a samsung evo 64GB sd card


    first sign of trouble -
    followed this article this and tried using ayufan's stable build 0.8.3
    the board quickly overheated and started rebooting....
    subsequent tries during the week resulted in pretty much the same, maybe with less overheating, but still crashes and reboots.



    found official build from the omv site - stable
    (or so it seemed)


    anyway - connected the WD drive first without hub and configured SMB and started using...
    figured performance sucked because of ntfs so reformatted to ext 4 and got the expected ~100MBps (on both read and write) and was happy...
    copied data from 2nd drive, connected it, formatted, copied data back...
    all seems to work great.


    2 days later I notice that suddenly any time I try to write a file it starts at full speed but a few seconds later speed drops to < 10MBps
    read performance still ~100MBps
    and basically NOTHING would solve this other than reflashing the SD card and starting over and the problem would ALWAYS return...
    I suspect it works till the OS crashes a few times


    I replaced the power supply, tried booting from USB instead of SD card
    tired without hub...
    tried various smb tweaks... nothing worked.


    Even when re-formatting the file system from OMV's interface it takes forever meaning OMV is slow to write on the drive LOCALLY, it's not an smb or network problem...
    I also suspected that it's bad contact on the usb ports reverting to USB 2.0 speeds, but that would kill the reads as well + writes still should be ~30MBps


    I'm beginning to think it's a faulty board + OMV's intolerance to crashes....



    Anyone experienced anything similar ?

    • Offizieller Beitrag

    I haven't figured out where the search button is on this forum yet so you'll have to excuse me...)

    Top right corner.

    OMV's intolerance to crashes

    This has nothing to do with OMV. Linux does not like crashes (as most other OS as well)

    the board quickly overheated and started rebooting....

    Are you using a heat sink? If yes you could try to reduce the CPU frequency.

  • Top right corner.

    Yes I kind of stumbled it after I finished posting and feel a bit dumb



    This has nothing to do with OMV. Linux does not like crashes (as most other OS as well)

    yes... I guess not, but not impressed by the inability to recover



    Are you using a heat sink? If yes you could try to reduce the CPU frequency.

    I have ordered some and they should arrive in a couple of days
    but I didn't necessary notice it crash because of heat...



    I noticed you have an odroid HC2, how happy are you with it?
    I thought if the rock64 won't work for me I might look for something else and the hc1/2 look promising
    maybe with a WD red drive which is like built for NAS..
    or other recommended drives ?

    • Offizieller Beitrag

    I noticed you have an odroid HC2, how happy are you with it?
    I thought if the rock64 won't work for me I might look for something else and the hc1/2 look promising
    maybe with a WD red drive which is like built for NAS..
    or other recommended drives ?

    I am quite happy with it.


    If you spin the drive 24/7 a NAS drive might be ok. If you intend to spin down the drive consider to use a desktop drive instead. Depends on how often you have to access the NAS.
    On my HC2 I use a 128GB SD card. Part of it contains a 3rd partition which I use for SMB to access frequently used files. So most of the time the HD is not spinning.

  • If you spin the drive 24/7

    The problem is NOT spinning it 27/7 because some drives don't support hdparm and therefore OMV's setting don't work
    so either get a NAS drive or a drive confirmed to work with hdparm


    there is hd-idle but it sometimes causes drives to unmount and not an elegant solution

    • Offizieller Beitrag

    On HC2 the spin down is done by the firmware of the SATA controller. No messing around with hdparm or hd-idle. Just works (you need to flash the firmware to set the spin down time)
    https://wiki.odroid.com/odroid-xu4/software/jms578_fw_update


    On my other server I use hd-idle. No issue.
    hdparm also worked to spin down the drive, but increased load cycle count extremely so I was due to reach spec limits very soon. Switched to hd-idle and load cycle count increases very slowly.

  • Can you possibly share your hd-idle config?
    I basically have a compiled .deb file I keep at hand so I don't have to create it each time..
    edit the default configuration file to start it automatically,
    set 10 min / 600 sec timeout on all drives
    and set a scheduled task in OMV to start it at boot because... not sure actually but the guide said it...

    • Offizieller Beitrag

    To enable hd-idle so that it is started at every boot:
    /etc/init.d/hd-idle start or
    systemctl enable hd-idle


    To check if hd-idle is running (e.g. after a reboot):
    systemctl status hd-idle


    This is a config file to spin down all drives after 10 min (all other lines are commented out with #

    Code
    START_HD_IDLE=true
    HD_IDLE_OPTS="-i 600 -l /var/log/hd-idle.log"

    this will also create a log file



    -l <logfile>
    Name of logfile (written only after a disk has spun up). Please notethat this option might cause the disk which holds the logfile to spinup just because another disk had some activity. This option shouldnot be used on systems with more than one disk except for tuningpurposes. On single-disk systems, this option should not cause anyadditional spinups.


    http://hd-idle.sourceforge.net/

  • Ok I spent the last 2 days tweaking this thing and I think I finally got it stable
    I made like 6 sd card images each "stage" I configured something and didn't break it.


    eventually I think the main suspects for breaking my writes are:
    1. temps... will get my heat sink and it will probably be solved.. maybe look for a fan
    2. s.m.a.r.t (idk why but after I turned it on the problem appeared shortly after)
    3. hd-idle on the Transcend drive... it would sometimes dismount it / cause the problem evntually


    eventually I configured hd idle like so:
    START_HD_IDLE=true
    HD_IDLE_OPTS="-i 0 -a disk/by-uuid/cf406f16-767a-4b74-a1c2-4ad375f15e12 -i 300"


    the id is of my WD drive
    didn't need to touch systemctl or configure scheduled tasks


    as for the Transcend - normal settings didn't spin it down but hdparm worked
    so I found this script and used it

Jetzt mitmachen!

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