Beiträge von DHGE

    syncthing takes a lot of time building up its indexes

    for 720 GB it takes on my ODROID ARM machines around ten hours on luks-encrypted volumes


    setting up a new server i start syncthing with no shared dirs

    if it is running (webinterface shows up) then I add a folder

    after this folder is synced as it should I add the other folders and let the server work...


    This might be easier if you already have some other machines syncing happily:

    Make sure (on the new server) that the prefix for folders is the right one and add your new server to another already working one and then enable sharing there. Then just hit "OK" when asked if to share on your new server.

    I think the m1 needs an external PSU for the SATA.

    For 2.5" the recommended power supply for these SBCs will suffice. For 3.5" you need 12V and 15+W extra only for the M1.

    Have a look at the wiki and the forum discussion.


    I am with you regarding the case of the HC4... That is why I hope my HC1 and HC2 will last long.

    The HC1 is running 24/7 since August 2017 here. One hard drive died and the data had to be recovered from backup.

    @adldcdk


    @alenrqz21

    I just see they have taken the HCx off their shop website...


    The N2+ is their fastest Arm-SBC I think.

    https://forum.odroid.com/viewtopic.php?f=211&t=44399

    I do not know this device - have an older XU4Q which I use as a backup with my even older external 3.5" USB Harddisks.


    For OMV/NAS I would always go with native SATA and not USB.

    Reliability wise I can only complain about my Raspis and that probably came form the poor power supply (even with a powered USB hub or powered USB disks).


    Since they seem to have discontinued the HCx:

    Today I would go with the HC4.

    I do not like the design: I would use duct tape to close the gaps since I only need a single 2.5" HD to keep the power consumption low.


    The just introduced M1 is better (NVME + SATA) and faster and more general purpose than the HC4 but I would wait until July 2022 or when I read about a first success with Debian Bullseye.


    For just tinkering I also would prefer the M1 over the N2+. But for tinkering I use my Raspis :)

    I do not know if you are attached to the Raspi(s).


    I would rather look elsewhere.

    https://www.armbian.com/download/

    https://dietpi.com/#download


    I switched to the ODROID/Hardkernel devices.

    For NAS the HCx (I have a HC1 and a HC2 as backup)

    They have newer devices but I like the design of my HCx. The latest M1 is tempting but there might be teething problems with the software right now.


    With going away from Raspis you get reliable powersupply, better cooling, better throughput for the disk(s) and often more CPU power(plex, nextcloud).


    4GB are enough - more is better. atm difficult to buy...

    I am in the process of updating my ODROID NAS devices to OMV6.

    I am doing a fresh install of Odroid XU4 / HCx Armbian 22.02 Bullseye Kernel 5.4.y Updated: Feb 27, 2022.

    I only use my backup of the configuration of Syncthing and Baikal but that is unrelated to the problem I describe here.


    So far I had good success with my HC2.

    Yesterday I worked on my XU4.

    As with the HC2 I installed the luks-encryption plugin via the webgui.

    Trying to decrypt the partition via webgui on the XU4 I get a "Gateway Timeout".

    Doing a cryptsetup luksOpen (or open) works on the console of this XU4 and is the only way atm to get the device running as a Syncthing NAS.


    The attached USB drive has two partitions:

    sda1 NTFS for Windows partition images

    sda2 luks encrypted ext4 for the NAS files


    On my HC2 the luks encrypted ext4 disk is attached to the (USB bridged) SATA port and has only one partition.


    The number of partitions and the type of connection is the difference I see here.

    I chose the individual DTBs (Device Tree, XU4 or HCx) via armbian-config though.


    Same Armbian image, same openmediavault 6.0.23, same apt upgrades.


    Having SMART in the dashboard also does not display anything on the XU4 - permanent "progress"bar running though.


    Decryption via webgui worked on my XU4 in OMV5 and works (with identical OS) on the HC2 in OMV6.

    Code
    /dev/disk/by-label/data2: can't find mount source /dev/disk/by-label/data2 in /etc/fstab

    check your fstab

    the mount sources could read

    /dev/mapper/dataX

    OMV does some mapping behind


    check it with ls, try to mount them manually

    if successful then check the non-working (in your case) fstabfor example with

    ls -l /dev/disk/by-label/

    here:

    lrwxrwxrwx 1 root root 10 Dec 18 09:32 WDred5p3T -> ../../dm-0

    which points to the first luks device dm0

    Here it works. I decrypt the drive after boot via the web. Then reread the shares via the MiniDLNA-refresh, restart Docker


    Maybe you have to do a mount -a after unlocking manually if you find no fault in your fstab.

    1) What is best practice if I do want to boot from the HDD in a separate partition?


    2) What's the performance with an Odroid HC2? Does it get significantly hotter because of CPU usage? Is the speed bearable for usage as a media folder that should also host a large photo collection with Lightroom?

    1) Why would you want this?

    My devices boot from SD-Cards, a XU4 also from EMMC

    2)

    - 4.6 MBytes/s when copying. 1 GByte in ~ 4 minutes transferred.

    - temperature insignificant during file transfer

    - sometimes I use my NAS ODROIDs to display fotos on my TVs, I would not use it for heavy editing (possible though)

    On my ODROIDs I run syncthing to keep files/directories synced between all my devices, foto editing is done locally and immediately synced after editing -> I do not care about the speed since it is done in the background anyway

    I had a hard time upgrading one system from OMV4 to 5.

    I totally messed up the second machine with the attempt of an upgrade and did a clean install.


    Part of the problems following ryecoaaron procedure in the thread 27909-omv-5-0-finally-out/&pageNo=2 was that when working as root the path is not amended for the system utilities.

    su ...

    /usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games


    I can cure that with

    Code
    export PATH="/sbin:/usr/sbin:/usr/local/sbin:$PATH"

    each time I want to work as root typing su and giving the password.


    The first lines of root .bashrc read:

    # This file is auto-generated by openmediavault (https://www.openmediavault.org)

    # WARNING: Do not edit this file, your changes will get lost.


    sudo su give me a suitable PATH.

    /usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin


    Is this a feature or a bug?

    I expect to have the same behavior than in Debian where I have the correct path for root and can customize roots .bashrc

    you might do a
    man cryptsetup
    once logged into your device via ssh


    this is what I did yesterday with a new external USB drive:





    this is how I mount them (not OMV but Manjaro) at boot via /etc/rc.local:

    Bash
    # falls es ein /dev/sdc1 (hoffentlich eine verschlüsselte USB-Platte) gibt: einbinden
    if [ -e /dev/sdc1 ]; then
    cryptsetup luksOpen --key-file /etc/keys/nc6400.key /dev/sdc1 backup_crypt
    mount /dev/mapper/backup_crypt /backup
    fi

    FWIW
    I have an ODROID hc1 that needs (maxes out at) a single 2,5" drive.


    When there is a sale on the large Seagate external drives like
    Seagate 2.5" USB 3.0 Backup Plus Portable Drive 4TB
    I buy one of these and break the case to get the drive since this is chaeper than byuing the same bare Barracuda.
    (Normally I use them the intended way and have now 3 of them as a spare for my hc1).


    These larger capacity 2,5" drives use SMR technology just like the WD Purple and most of the larger Reds.


    They work for years here and I am on a GBit network (-> 100 MBit/s under ideal conditions) and do not care too much about how long the stuff is running in the background.
    Reads and writes are mostly run via Syncthing here.


    TLDR: Purples are just fine and fast enough.

    I wonder why it is still not fixed...


    I just setup a new ODROID hc2 and DLed the latest image from sourgeforge and ran into the same issue.


    fix: (as root)



    Bash
    wget -O /usr/lib/python3.5/weakref.py https://raw.githubusercontent.com/python/cpython/9cd7e17640a49635d1c1f8c2989578a8fc2c1de6/Lib/weakref.py

    I had to start over with etching an SD card since the unfinished dpkg -i process made my ssh logins impossible.


    OOPS: seems to be still in the repos
    After a dist-upgrade on my fresh install I hade to fix this again

    I see several ways:


    - what makes the mount disappear? find the root of this and a remedy
    - use rsnapshot (a sophisticated script with exceptions, generations, ...) it does not start without the proper path mounted
    - write a batchfile that checks for an item e.g. a directory that you know must exist on your drive and do not start if the check fails
    - look into automount (overkill IMO)


    btw. if I use rsync for backup I use the "--delete" option