Posts by Rd65

    i am happy with my new m5015 (LSI 9260-8i)... i use 7 sata disks a 2 TB. Next time i may use a LSi 9266-8i but this one was cheap and i can't use pcie3. the more important question is which disks are used. 2 or 4 TB wd red drives are ok but shingeling drives like seagates are verry bad! so its easy to get a fast and reliable 25 tb storage (8*4 tb)... but 50 tb.. hmmmm... ok may be with 2 controllers, 2x 8xpcie onboard and a huge cassis with 16 drivebays... but if you plan this... then a SAS Storage Array Expander and SAS Drives may be a better choice. and it even makes no sense to do softwareraid on 16 disks...

    50 Tb... wow... how do you connect these Disks? Do you own a SAS Storage Array Expander Rack?
    i think its better to start with a 5TB Array... and some Sata Disks...



    maybe you searching for something like this:


    Evtl. mal eine andere usb-sata bridge an der ssd versuchen.
    Es sollen sich da sehr instabile Teile im Markt befinden.
    Den USB-C Bug würde ich mal ausschließen - Spannungseinbrüche am Netzteil jedoch nicht.
    Ein 15W Netzteil braucht der Pi schon selbst, versogt man noch eine externe SSD ohne eigenes Netz, brauch man min 20W und hat dann noch keine große Reserve. Die veranschlagen sogar 7,5W für eine SSD.
    https://www.heise.de/ct/artike…Alleskoenner-3836662.html

    is there a difference between the motherboard(on first post) to others on the market? Should I pick different motherboard?Which GPU should I pick if AMD CPU doesn't come with integrated graphics?

    yes. it depends on which additional hardware you plan to use.

    maybe. see above


    there are lot of amd boards out there with grafic enginge. but VMs do not need a Grafic card.


    if you want stop posting me, ask me... but you want help.. my nas running fine including actual 2 VM's. :)

    hm.. ok i will try the omv-nginx plugin again with my fix to the omv-webgui site. As you told, it may work like expectet. i don't test that till now.


    yes docker may be easy to handle... but i don't download and execute code that is unproven, depending on unknown people and complicated to overview. its worthless to use docker for things which i can manage easily native. lots of people may do that... but not me. This is more a philosophic thing, we will not find a common Opinion about that.


    We talked about the development and designrules of omv... unplesant.

    i found a solution ... the problem depends on a misconfig of omv web itself.
    1. we disable the omv-nginx plugin.
    2. we enable the default web page with: #nginx_ensite default
    this will cach up all web request not fitting a virtual host name.
    its webroot is /var/www/html so we can place a empty index.html or use the default nginx webpage with a short greeting.
    if we now restart nginx with #service nginx restart the server will fail.
    this depends on a wrong config in /etc/nginx/sites-available/openmediavault-webgui
    this is the omv admin webpage. we will fix em.
    3. we rename server_name openmediavault-webgui; to the name of the server as defined in hostname. in my case nas.

    Code
    # server_name openmediavault-webgui;
    server_name nas;



    now the Server will listen on name nas. means a http://nas show us the omv page.
    nex part to fix is the port definition.
    the virtual server nas uses:


    Code
    listen [::]:80 ipv6only=off;
    listen [::]:443 ipv6only=off ssl deferred;

    4. this breaks the rules for virtual server definitions. we change that to:

    Code
    listen 80;
    listen [::]:80;
    #listen [::]:80 ipv6only=off;
    listen 443 ssl deferred;
    listen [::]:443 ssl deferred;
    #listen [::]:443 ipv6only=off ssl deferred;

    it looks like the same but is different.
    now you can start/restart the server with a cachall default site and a virtual server omv page on the same ip with name nas.
    #service nginx restart
    now you can add more virtual websites with ddns servernames, other names and so on.
    would the web plugin work, it would now possible to adjust web pages with virtualnames . But omv-nginx can't handle multiple sites on the same port. but now its possible to do it by hand like this.

    Code
    server {
    listen 80;
    server_name my.ddns.server www.my.ddns.server;
    ...
    }

    to protect these change against updates and wrong modifications, you can fix this by setting the immutable-bit:
    #chattr +i /etc/nginx/sites-available/openmediavault-webgui
    if you want allow changes, remove it by #chattr -i /etc/nginx/sites-available/openmediavault-webgui


    read carefuly omv update anouncements if they fix this problem.
    always backup old states and your changes. maybe omv-firtsaid will reset your web config.
    so doublecheck your webconfig if you save things depend on nginx.
    hope that will help to fix this issue for others.


    i think it is a real bad idea to run a webserver with php7 maybe exposed via ddns without a cachall site!
    this invites bruteforce hackers and scriptkiddies.


    and if someone is able to review or fix omv-nginx portcheck, please do it... now it is it worth. maybe its working now with these modifications but the thrown errors looks (to me) like a problem in omv-nginx too.


    a last tip, if you want use https with default server via ddns, place the snakeoil cert in the default server.
    there are scripts outside reading cert info and use emails and hostnames as startpoint gathering infos about you.
    Use your letsencypt cert only in virtual server_name wich is signed for. someone who didn't know your ddns name is not able to ask the ddns server something.

    Raspis die unter Last laufen werden meist wärmer als sie dürfen. Eigentlich sollten sie dann runter takten.
    Es wäre möglich, das dies zu Kommunikationsproblemen führt weil ein USB3 Bus eben doch ordentlich Rechenleistung braucht.
    Das kann im Zweifel auch bedeuten, das der Raspi die Partition kaputt schreibt - z.b. nach Bufferüberläufen usw.
    Das wäre ein denkbares Scenario. Ob das so ist, müsste man jedoch genauer untersuchen. Z.B. mit Kühlspray. während des Kopiervorgangs als eher handwerkliches Mittel um einen Anhaltspunkt zu bekommen.


    Ansich behandelt Linux SSD Platten und SATA Platten gleich. Allerdings haben Platten generell unterschiedliche Features wie z.b. den Trimm Befehl oder Command Queuing. Mit libata.force=noncq im kernelstart lässt sich z.B. das Command Queuing abschalten. Es geht aber auch per sysctl. Alte SSDs haben auch gern schon mal Probleme mit ihrer Firmware.


    Zudem kann man auch mal über die shell in die Logs gucken #dmesg und #iostat oder #iotop mal schauen was der Rechner so treibt. #tail -f /var/log/kern.log oder #tail -f /var/log/syslog schauen beim kopieren evtl. auch.
    Ohne genauere Angaben ist eine Diagnose schwierig bis unmöglich.

    Ok.. i found the next bug in omv... :) exactly omv-nginx.


    2 sites on one port without name support is not possible , they are only port driven. thats clear.


    2 sites as different virtual server names on the same port are usual... but omv-nginx complain about previously used ports if you try to save the second virtual server name. there is an extra switch in the configweb to change from port to virtual name support... so he has no reason to complain if the servernames is different... but... he complains.
    thats wrong behavior.


    additionaly i try ipbased virtual server (means eht0:1 and eth0:2 with differen Adresses) but in the omv-nginx you can't setup ip adresses for servers... but this is not a bug, more something like a "missing feature".


    i am realy wondering that nobody else complain about that. This bug means - no native virtual webservers on omv... this makes the omv-nginx plugin worthless. but please.... dont treat me again with... use docker, use webmin, use whatever...

    I need some help or advice configuring nginx. i know apache2 configs but nginx is new to me.


    My setup:
    ISP--<dsl>--Fritzbox with ddns--switch--nas(omv) and other clients
    the fritzbox may route ipv4 and ipv6 ports to nas.


    my nginx on omv-nas (default server) shoud be accessible via http and https only from internal net.
    but it shoud serve webpages for some external access (via namebased servers?) too.
    so we have different scenarios for different Sites.
    These are possible urls against nginx


    nas -> intern
    nas.local -> intern
    192.168.178.111 -> intern
    my sqlmanager site -> intern
    my bittorrent manager site -> intern
    78.1.2.3 which may be my ip4 adress -> extern error page
    my.ddns.host -> extern Site
    2001:4312:.... which may be my ip6 adress -> extern error page
    even blind port access without a hostname routed by fritzbox -> extern error page
    the webdav page... -> intern & extern accessible


    i have installed the omv-nginx plugin too and able to config additional namebased sites prepeared with letsencypt..
    i think this config depends on the "default server" and i need to config omv site on a namebased server.
    but there is no way to config that. the omv-site seem to be always the default, plus additionaly configured Sites in nginx plugin.


    But i dont want to show omv loginpage on all not-namebase-fitting web requests from outside... i want to send those brutforce clients to a error page.


    On apache i would config a default server with a error Page and 2 namebased servers as nas and my.ddns.host.


    Does someone have a hint or a howto link?

    At the end, it is your job to fix always possible upcomming FS errors and search for weaknes in your NAS. Not the Guys which told you fairy tales about something they never used by own.

    take al look of reasons why disks fails and data get corrupt.
    if ram randomly fails, no filesystem on the world is able to correct that.
    But a smaller footprint of ram reduce the chance the filesystem is involved in Memory errors. logic. at the first view.
    To check if ram fail, there is ECC... a nas without ECC Rams maybe a more risky constellation.
    Mechanical Disks get often to hot. The CPU too. if the Disk Temperatur gets over 70 degree C...the risk of some data loss is 100%. even 50 degree C are to much. But Disks use checksum error correction inside. And smart take a look on that. If Smart complain, this has reasons! To eliminate Errors they invented Raid ... hardware raid... sofware raid.. with 1 controller, 2 controllers... raidlevels.. and so on. Doing mirroring 2 Disks on the same Controller/bus is simply a wrong aproach. And in highend Servers yo will never find such bad design bug! Good mirroring needs 2 independent controllers!
    A good strategy against Dataloss are backups, md5 checksums on files, a well planed Data path, raid, a stable filesystem and a good Hardware plan. dont pick one of them, all are important.
    i personaly prefer ext4 (a journaling system) with tools to check FS consitency start on boot and per cron.
    Others prefer XFS... i think its a good one... but im not expirienced in xfs. ZFS may work too.. but its known as memory consuming. All Filesystems and all Hardware work as expected without errors if you use them as intended,
    If this Basic isn't ususal, then put your raid in the trashcan and try to build a new one.
    These Aspects are chained togeher and the weakest Chainring will break if it works outside its specifications.
    I predict that it has the least to do with the filesystem. I think it rather a matter of taste which Filesystem you use.
    best use one that you are familiar with and that you can deal and trust with

    i'm happy to see, that you are doing progress.
    yes you can try it in the command field.
    update 1:
    its difficult to use html output, as i told. its possible but its step 387... you are on step 2 or 3 on the learning curve.
    but dont give up...
    a way may be change all /lf chars (linefeed) against a <br> sed (streaming editor) can do that.
    update 2:
    yes tree is nice. there are lots of more possibilitys.
    "." is your startpoint "/" is the root of the tree.
    maybe take al look at
    tree -DthH "/srv/diskname/sharename/"
    In a shell you can use a tab key to list alternatives...
    #ls -la /srv/[tab] will show you diskname1, diskname2, ... its eayser no navigate and you dont need to remember all names.
    thumbs up!


    edit: ah no.. tree uses html output with -H. fine.
    i dont use tree but yes... maybe this help you to output html.
    maybe you can use a find command like this:
    #find /srv/diskname/sharename/ -type d | tree -DthH
    means... find al dirs starting at your shares and use tree on it.
    this can be done better... but for learning it is the next step

    ok a verry short and fast briefing to use a shell,cron,bashscripts... on pi:
    Use strg-alt-f1 strg-alt-f2 ans so on... to get a getty shell. use 3 keys together. most times on strg-alt-f6/7 you are back on X.
    the strg key may be signed als cntr or what ever ... strg is the german key layout. on pi and other nas:
    you can use most times ssh ... use putty on windows to login via ssh.
    or install omv-plugin shellinabox and do it via webpage.


    so try it on the shell: you will see.
    the ls command as example do the file-job and a execution via cron or anacron do it timed.
    to do this automatic, you can use plugin omv-anacron or use scheduled job (cron) to execute it on time. its a good start to learn about cron. on shell you can use pure cron/crontab or the cron-dirs in /etc/. edit text with #nano /textfile.txt
    try the forum search to find howtos about these plugins/tools and how they work.
    on shell, you can use #man cron and #man crontab to get more info. the web is full of hints. use q to leave man.
    you can edit a bash script like this on the footer.. without numbers...namend myscript.sh, put it in /etc/cron.hourly and your computer will remind you once per hour with a short beep-sound. maybe you need a speaker too... make it executable and the fun starts :) chmod a+x myscript.sh In this way you can do more shellscripts... but be carefull... you can easyly destroy things on shell. Use backups on your nas! If you hear the beep once per hour, you can exchange beep against ls and do text output to your wanted target dir at every time you want. Use this as a challenge to learn about cron...try to config cron to exactly beep on every 15 min beginning exactly at the hour like a real clock. cron wil do it timed or never, anacron wil remeber lost events (if a event ocure and your nas is off) and catch up the action. so beep with anacron is a bad idea... backup with anacron may be good. you can test the script by hand... use #./myscript.sh and remind the ./ at start. your script will not start without that ./
    if you are in panic, try #exit to close the shell. but its better not to panic....


    • #!/bin/bash
    • sleep 1
    • beep

    ok i try to answer.


    we think you want to buy a car.
    you need a fast car. for whatever reason. build in Germany. becaus of no speedlimits on Highways. maybe...
    but you also want to make apartment moves so that you can save shipping costs.
    it should have nice alloy wheels
    and pull at least as much as a john deere tractor
    in the mud...
    and it can not cost more than a month's salary .. no one knows how much that is


    That's about your requirement profile... translated to a car.
    the problem is ... you can buy everything individually. maybe cheap. But not together, and if it does, then it is very expensive.
    Its the same at your requirement profile.


    fast encryptet filesystems need special hardware which is expensive. Let's talk about 6GB/s SAS controller with lots of cache.
    A VM with windows needs lots of Ram and fast CPUs and fast Disks. but fast cpus and disks use lots of power.
    Now we have a Server with an SAS, some Disks (raid is a mini form of encyption too, but you can use expensive encryption keys)
    A normal Nas uses 20W-150W/h... your Server... maybe 600-1000W/h.
    But does it fit youre requirments?
    Its big, its fast, you can use VMs... but its not a home-nas and it costs 10.000€!
    Your design is a compromis and it will work but noboddy can answer, if it is a good one.
    You can size your needs and size the hardware, but both depends on you.


    Lot's of People uses spare parts for nas, build them, try them, change parts and so on. this is why its called homegrown.
    The bandwith goes from RasPI with old USB2 Disk till Rackserversarrays... depend of the User. There is no "right one".
    For Beginners the easyest and cheapest way is to start with a Raspi3 or 4 for about 50 Bucks...and a usb3 Disk. Its equivalent to small nas entry models like synology and qnap. Some People use 2x8TB on these Dwarfs.
    But if you want P.O.W.E.R ... your design may be ok as a entry solution. but you will curse on the Energy costs.
    I think its more important that a System is stable and reliable.


    I use a AMD Fx8150 with 16 GB Ram and sata softraid .. it fits my actual needs, but most of time it sleeps. If it dosn't fit anymore, i will build the next one. smaller.. bigger.. who knows... A nas ist not like a Car... if you put a Raspi in the Trashcan, or use it for other projects... or reassemble the big nas for other needs... noboddy else can tell you if it is worth.