Posts by puterfixer

  • HDD failure - disaster recovery

    I'm a linux noob, but from what I just googled it seems that linux normally runs a fsck at start-up, and that's what triggers the console prompt to continue. Apparently you could edit the grub bootloader configuration file and append the fastboot parameter to the kernel options, in order to skip the initial fsck check. In addition to this, you'll have to come up with a way to send an e-mail with a warning afterwards.


    Or, you could go on eBay and buy a hardware RAID controller which does a lot of neat things, and by-pass the software RAID completely. I'm just looking at a pair of HP SmartArray P400 cards with 512 MB RAM and battery back-up, on PCI Express x8, which support up to 8 SAS/SATA drives each, and they're going for $90 for both. The only risk for this set-up is that, if the hardware controller fails, you'll need a spare one to retrieve the data, as opposed to the software RAID which can be rebuilt on any other hardware.

  • Problems suspending/resuming

    bellyman, for a PC there are several levels of power saving, the deepest one being with everything turned off except for the NIC which is still listening for traffic. Now, in order for the system not to power on for ANY kind of traffic, including stupid broadcast messages which are not intended in any way for that system, there's a "key" to trigger the power on remotely, and that's the magic packet.


    But this is not implemented as an automatic feature in any operating system, as you incorrectly assumed. A remote system may become inaccessible for a number of reasons, not just the power management, so it doesn't make sense to try and send a magic packet. Also, a remote system doesn't know and cannot determine the MAC address of the server; this is available only on the same network segment, and not beyond. So it really is up to a system admin to remotely wake up the system KNOWING the wake-up key.


    A system which is supposed to be available at all times remotely should never be allowed to get into such a deep power saving state to begin with. There are shallower levels of power saving in which the network card wakes up the PC from a sleep state for ANY data traffic, allowing it to determine whether the traffic was a legitimate request or should be ignored. Standalone NAS systems do this - they can afford to with their low-power specialized chips anyway, as opposed to a more powerful, more versatile computer.


    Also, get a watt-meter to determine the real power consumption. Some people assume the maximum theoretical debit of the power supply is actually how much it draws from the grid, which is grossly incorrect. My Seasonic 380W 80-plus certified power supply used to draw 80W on the old, power-hungry system in idle with 4 drives running, peaking up briefly at 130W in full load.

  • Problems suspending/resuming

    The so-called "magic packet" which triggers the WOL feature of the network card contains the MAC address twice. But, since the target system is essentially powered off and doesn't have a working TCP/IP stack that would allow network connections, you can't use TCP protocol and you can't use the target system's IP address.


    You have to use UDP, which fires a packet blindly without opening a connection first and doesn't expect a reply. Also, you have to use the subnet's broadcast IP address so that the packet is relayed to all systems in the subnet; this is usually the last IP of the subnet (i.e. 192.168.1.255). And, of course, you must know the MAC address of the system, which will be assembled in the magic packet.


    == LAN wake-up ==
    All you need is a tool which easily broadcasts the magic packet in your network to wake up the target PC.


    == Internet wake-up ==
    If you also want to send the magic packet from the Internet through a router, then you must configure the router's "Virtual Servers" section to forward all incoming communications on UDP port 9 to the subnet's broadcast IP address, also UDP port 9.


    If possible, also subscribe to a dynamic IP service such as dyndns or no-ip and configure the router to use this service. This way, every time your router connects to the Internet (DSL or PPPoE or whatever) and gets a possibly different IP address, you can still access it with a personal hostname which the router's client updates to its new public IP address.


    To wake up your PC from the Internet, you either use a program on someone's computer to send the magic packet to your router's hostname (via UDP port 9), or you use some web-based service to do this. My favorite is dslreports.com/wakeup .

  • Install OMV on a Compact Flash card.

    Iro, my initial setup was on a CF card mounted in the IDE slot. Unfortunately, after a couple of months the system started hanging unexpectedly. As soon as I migrated the data from the CF to a standard IDE drive, the problems disappeared.


    Strange enough, the CF card doesn't seem to have any defects - I ran a quick test on it. But there might be something wrong there that I didn't discover yet.


    CFs are also Flash chips, just in a different package/interface than USB (and usually a lot faster, too). They suffer from the same risk of burn-out, unlike SSDs with TRIM function. And consider that OMV is frequently modifying files on the system disk - at least the log files which grow constantly, and the traffic/load/etc. graphs which get overwritten every 5 minutes. Yes, most likely the same memory cells will be used over and over and over again, until they burn out after some thousands of uses.

  • lcd4linux

    Thanks for the tip! I installed Mobile PC Monitor and wrote a newbie's tutorial on how to do this step by step. Waiting for approval in moderation queue now :)

  • Mobile PC Monitor

    Tomte mentioned in the lcd4linux thread an interesting remote monitoring/management solution called Mobile PC Monitor.


    I do want to have remote management access to my OMV, but without exposing the management interface to the Internet through the router. Also, the OMV Web UI doesn't like very much the constraints of my mobile phone's small screen. Nothing fancy, just to be able to restart or power off the system remotely, and check some "vital signs". Mobile PC Monitor seems a good solution for my requirements, so I decided to give it a try. Here's how I did it.


    INSTALLATION


    #1 - You need to install Java. Easy peasy, fire up the console as root and:

    Code
    apt-get install default-jre


    When this ends, you can check the version with the command:

    Code
    java -version


    Mine says:

    Code
    java version "1.6.0_18"
OpenJDK Runtime Environment (IcedTea6 1.8.13) (6b18-1.8.13-0+squeeze2)
OpenJDK 64-Bit Server VM (build 14.0-b16, mixed mode)


    #2 - You need to download the Mobile PC Monitor "agent", extract it and run the installer. Here's what I did in the console:

    Code
    cd /usr
wget http://www.mobilepcmonitor.com/download/pcmonitor.tar.gz
tar zxvf pcmonitor.tar.gz
cd pcmonitor
./install


    The installer will take you through a text-mode wizard, asking you to answer the following questions:
    - installation path: Enter to accept the default (/opt/pcmonitor)
    - user for PC Monitor: Enter to accept the default (root)
    - group for PC Monitor: Enter to accept the default (root)
    - confirm the installation with above parameters: Y and Enter (default is N)
    - start the PC Monitor Manager (requires GUI): N and enter (default is Y)
    - do you want to connect to an Enterprise PC Monitor Server: Enter to accept the default (No)
    - do you have a PC Monitor account: y or n, depending whether you have an account or not; I chose N and proceeded with account registration
    - username: select an account name for you
    - e-mail address: enter a valid e-mail for your account
    - password: choose a password for your PC Monitor remote access account, must be between 10 and 25 characters long
    - full name: self-explanatory
    - organisation
    - activation code: get it from the confirmation e-mail


    After this, PC Monitor agent is starting. The installer didn't do anything afterwards, but didn't return to a prompt either; I waited for a while, then ended it with a Ctrl-C.


    USAGE


    Here's how you can manage the daemon (agent) from a console:
    - to start:

    Code
    /opt/pcmonitor/bin/pcmonitor start


    -to stop:

    Code
    /opt/pcmonitor/bin/pcmonitor stop


    - to restart:

    Code
    /opt/pcmonitor/bin/pcmonitor restart


    -to check status:

    Code
    /opt/pcmonitor/bin/pcmonitor status


    The installation configuration file is saved in /opt/pcmonitor/conf/config.xml - in case you want to back it up for future reinstalls.


    To manage the systems remotely, access https://my.mobilepcmonitor.com/ or get the applications from Apple App Store, Google Play, Microsoft Windows Store, or the PC Monitor Dashboard for Windows (x86 or x64). Complete list is here: http://www.mobilepcmonitor.com/downloads


    Note: PC Monitor service is free for up to 5 monitored systems. If you want to remove one, you have to turn off the daemon first, wait for it to show up as offline, then you'll have the Delete option. My account says that I have a trial account for 2 weeks, let's see what I need to do afterwards.


    Hope this helps.

  • Special Network-Setup?

    Hah, I know how that goes - I've got about 1 TB of raw photos so far as well. But I'm not that paranoid about security; haven't had a virus or security breach in YEARS. Easy:
    - router filters out intrusions;
    - don't install crap (malware or pirated software from dubious sources) on your Windows system;
    - keep Windows up to date with security patches and updates (Microsoft releases monthly patches on Patch Tuesday = second Tuesday of every month);
    - Microsoft Security Essentials is sufficient as a realtime protection; from time to time (monthly) double-check it with TrendMicro HouseCall (free, http://housecall.trendmicro.com or housecall.antivirus.com)
    - use OpenDNS.org servers (208.67.222.222 and 208.67.220.220) to protect all your systems from accessing malware sites and botnets - define them in router's DHCP server


    If you want to test software, virtual machines are one way of doing it (I love VirtualBox), but it might be easier with Sandboxie (.com) + Windows' built-in System Restore as a backup.


    But, bonus points for you for going the extra mile with a separate machine for the web, and a Linux one too!

  • Trunking

    :D As I said, just checking. You'd be surprised how many people try stuff they don't understand, then wonder why it's not working. To quote a good line from a crappy Steven Seagal movie, "Assumption is the mother of all f***-ups" :D


    What cards do you have? Are they overheating? I modded mine with chipset heatsinks glued to the processor (Arctic Silver Adhesive, two compounds combine in a glue that is also a thermal conductor). Second point of failure is the optoelectric coupler between the network chip and the RJ45 port, the cheap ones tend to fry in high traffic.

  • Special Network-Setup?

    :D I'm still not so sure why would you bother with having a separate network with no internet access, when you are behind a router. That router is an excellent protection for intrusion; you actually have to explicitly define rules as exceptions to the block-all filter. You could keep all systems on one local network behind the router; perhaps, if the router allows, assign static IPs to known MAC addresses in the router's DHCP section, and even block internet access to specific IPs. It all depends on the capabilities of the DSL router.

  • Trunking

    Just checking - link aggregation requires the peer device (network switch) to also be compliant with LACP (Link Aggregation Control Protocol), otherwise you'll end up with a nasty loop in your star network and lots of unexpected results. I've seen link aggregation available only in managed switches, usually for enterprise use. Is yours compatible?

  • Special Network-Setup?

    Hey Dennis,


    From what I understand, you will be operating two distinct networks. One will be a private network between the host system with virtual machines on NIC1 connected to OMV and possibly other clients via a switch, with no internet connection. The second network would be between the virtual machines through NIC2 and the DSL router, which has internet connection.


    Each network can have a single DHCP server, if you don't want to manually configure IPs etc. On the second network, the router's built-in DHCP server will assign IPs to the virtual machines. On the first network, you'll need a separate DHCP server, unless you want to assign IPs manually.


    Note that if you physically combine the two networks in the same switch, you can still assign two completely separate IP classes to them. You'll have the router's DHCP server running, and the clients who are set to receive IPs by DHCP will also get internet access (such as the virtual machines). Separately, for the systems you don't want internet access, you can manually assign a different class of IPs with no gateway and no DNS.


    Even that they are physically on the same network, the systems with IPs in different classes will not be able to talk to each other, unless you set up IP aliases for the same network interface (multiple IPs on the same network card; Windows XP/7 knows how to do this).


    This makes it easy to switch between the two networks, by allowing a system to receive IP from DHCP for internet access or setting the IP manually to be in the private network. And, since you'll only have a handful of systems, it's not that difficult to ocasionally make the change for half an hour to get Internet access again.

  • Special Network-Setup?

    There can be only one DHCP server on a network segment at any time. The network administrator decides which system acts as DHCP server, to allocate IP addresses and announce the IPs of the DNS servers and the IP of the gateway. Usually this is the system which is always on. There is no mechanism to set-up a backup DHCP server of sorts.


    I'm not exactly sure I understand why would your router be inaccessible at times, so that the computers on the local network can no longer benefit from its built-in DHCP server. However, if you can get a DHCP server installed and running on OMV, you could configure your network as follows:
    - DSL router: disable built-in DHCP server and define a fixed IP address, such as 192.168.1.1, so that you can refer to it as DNS and gateway; check what's the subnet accepted by the router configuration and use that one;
    - OMV: define a fixed IP address (it is normal for the system with DHCP server to have its own IP manually assigned) - for instance, 192.168.1.2; then configure the DHCP daemon to give IPs in a small range (for instance 192.168.1.11-192.168.1.50) and announce the router's IP address as DNS and gateway.


    This way, the computers will always get their IPs from the OMV's DHCP server and be able to communicate on the internal network without issues. Their Internet access will depend whether the DSL router is turned on or not, with no additional configuration. And, you will have some IPs (.3-.10) reserved for manual assignment to other systems which you may need to reference directly by the same IP address every time, such as a network printer or a network camera or whatever.

  • OpenMediaVault Strange Network Activity

    Try turning transmission off (not just turn off all downloads) and see if that changes anything. From my experience, enabling DHT floods the network with so much (UDP) traffic that some routers can't keep up with the number of parallel connections. Keep in mind that SOHO routers have some limits in number of concurrent connections; some are as low as 32, others may have 500. You won't find this anywhere in the technical specifications, it's the torrent which reveals such a limitation first.


    On the other hand, ping is not a reliable tool to test throughput or bandwidth. Are you experiencing any other kind of issues or limitations when your OMV is running, or is the ping reply the only concern?

  • Very slow network speed

    Wireless is slow not necessarily from a throughput standpoint, but from a latency point of view. Also, it is susceptible to interference from other wireless networks nearby, operating on the same channels. 802.11n, especially the 300 Mbps kind, needs to have clear radio spectrum over several channels to work great. With overlapping wireless access points (b/g/n, doesn't matter, they use the same frequencies), there are frequent interferences and colliding packets which need to be repeated. As a result, you can experience more frequent interruptions, errors and a slower data throughput.


    For the wired connection, I'd recommend that you disable the IPv6 in Windows 7; leave only IPv4. My router doesn't get along with IPv6 at all, and it locks up if I try to use jumbo frames on its built-in gigabit switch, so I have to stick to small frames and their less efficient overhead-to-usable-data ratio even on a gigabit network.


    Set manual IP address to OMV if you have selected DHCP allocated IP by any chance.

  • OMV stops responding to remote access

    :D That's alright, all in all it took more time to move the box in front of the TV and hook it up to a USB CD drive etc. than to repartition everything. A few minutes after trying things I figured I don't actually need any of the data on the swap partition, it just needs to 1) exist as the first logical partition in the extended partition and 2) have the correct partition type. Click-click and it was done.