TurnKey FileServer suddenly stops the wired network interface

Hello All, I am using the version "turnkey-fileserver-11.2-lucid-x86" on a somewhat experimental setup, for me at least. I wanted to test turnkey in the most light-weight and frugal environment. I had a couple of old HP t5530 thin client lying around.

You can check its details over here.

http://www.parkytowers.me.uk/thin/HpT5530/index.shtml

This small machine has fixed 128MB RAM and I had a 512MB Flash DOM installed in it for storage. Although i did read somewhere that the turnkey team has tested the server in the most minimum setup of 256MB RAM, but I wanted to go below and beyond by trying the installation on this little guy. So i burned the ISO file on a USB flash stick using the unetbootin from http://unetbootin.sourceforge.net

It makes the USB stick bootable with default GRUB screen of running the TurnKey from USB or installing it onto a harddisk. I chose the first option of running it from the disk, and after waiting for quite sometime (have to give the discount for miniscule RAM and old hardware) it showed me the standard options of selecting IP address and stuff and finally started the turnkey server. I could check it by connecting to this machine from another one at the default https://myipaddress:12341.

This was a good news so I went on by trying to install it on a harddisk, but since the internal 512MB is too little , I added another flash USB stick (4GB) to the thin client to be used as installation disk (to be booted from USB). I chose the manual partitioning and kept all the partitions on the USB Flash stick except the /var partition which i thought (might be for frequent access) should be kept in internal 512MB DOM disk.

The idea has worked and now I have a running TurnKey Fileserver on HP t5530. I have connected two USB harddisks (2TB each) to it and shared them via SAMBA networking. Have installed NTFS-3G in TurnKey Fileserver so the disks around mounted in fstab in read/write mode, somehow the standard ntfs filesystem built-in does not support SAMBA write on NTFS shares. The read/write rate i get on my CAT5 10/100MBps wired network is around 4MB/sec. Not very fast but enough for the while. I can stream the shared movies from the SAMBA shares to my NMT and XBMC based HTPC. The 1080p stuff streams without stuttering as well, more or less.

Have updated the ubuntu packages as well and tried the upgrade distribution option too. Everything works except that a major glitch. During reading/streaming or writing to the SAMBA shares, sometimes at totally random time intervals, the WIRED network connection to the turnkey server drops. The network adapter totally stops responding. When this happens I cannot ping the turnkey fileserver niether can I access the SAMBA shares or the WebMin interface. Seems like there is no network adapter present in the machine. The only solution to it is to physically turn-off the machine and turn it back on after which everything starts working normally untill the same network black out happens again.

I am no expert in CLI Linux, whatever the underlying Ubuntu has configrued is pretty much plug-n-play for the system devices. The network adapter as far as I know inside the machine is RealTek. You can check the datasheet here http://www.forestals.com/filebank/imagebank/products/T5530.pdf

Is there any way i can remedy the situation for a stable wired network connection, or maybe a way I can do troubleshooting or can get some logs to post over here for further guidance. Thanks for reading the long post and replying in advance.

Replies (4)

Will check and get back

Reply 1 2011-10-06

Right now the machine running TurnKey is connected headless to the network router.

Will try connecting it with display. So if it does respond normally with a dead network interface, i will run the command you mentioned and will post the output.

Thanks.

Do you have a screen connected?

Reply 2 2011-10-06

If not perhaps it is not just the networking that stops?

If you do have a screen connected and the rest of the system keeps working then you could do some troubleshooting. Firstly I'd check what the kernel is up to tail /var/log/dmesg (or just dmesg to look at it all). may give you some idea. Other log files are found in /var/log It's a complete stab in the dark, but maybe it's worth trying to update the driver to the latest one from the Realtek? (I wouldn't bother trying that until you confirm that it definately is the NIC that is dropping out).

Got the hang of it, without a solution

Reply 3 2011-10-07

Hello,

I have connected a display and keyboard directly to the previoulsy headless thin client running Turnkey, acting as a server for my network.

Tried connecting two harddisks mounted with NTFS-3G and SAMBA shares available on network. Did various big >4GB file movements of reading and writing on the shares.

The same loss of wired network interface happened again , just when it happens. the tail var/log/dmesg shows the following:

root@fileserver ~# tail /var/log/dmesg
[ 10.717561] shpchp: Standard Hot Plug PCI Controller Driver version: 0.4
[ 11.654052] parport_pc 00:09: reported by Plug and Play ACPI
[ 11.654147] parport0: PC-style at 0x378 (0x778), irq 7, dma 3 [PCSPP,TRISTATE,COMPAT,ECP,DMA]
[ 12.125214] ppdev: user-space parallel port driver
[ 12.592290] eth0: link up, 100Mbps, full-duplex, lpa 0x45E1
[ 13.026555] ACPI: PCI Interrupt Link [ALKC] enabled at IRQ 22
[ 13.026580] alloc irq_desc for 22 on node -1
[ 13.026591] alloc kstat_irqs on node -1
[ 13.026619] VIA 82xx Audio 0000:00:11.5: PCI INT C -> Link[ALKC] -> GSI 22 (level, low) -> IRQ 22
[ 13.026836] VIA 82xx Audio 0000:00:11.5: setting latency timer to 64
root@fileserver ~#

During this while the wired network interface is down. So I swtiched to the KVM connected to the server machine physically. Usually from a cold boot the machine boots up and reaches the following screen and stays there.

/screenshots/configuration-console-usage

In the above screen under normal running condition the cursor is blinking at "A" of the "<Advanced Menu>" option as shown.

However during a network blackout, the cursor was soild (not blinking) at the same position as described above.

As soon as I pressed any key on the keyboard directly connected to the server machine, the cursor started blinking again and the lost wired network interface was restored. The ping response from the connected Windows machine showed the signs of life again.

I have tried it twice and it behaves the same way. A rather unlikely and unexplicable sort of bug/behaviour which is beyond my understanding. Very strange and weird. It maybe happening at regular fixed time intervals, but to me it appears rather randomly during the read/write operations to the connected harddisks.

THe server is usually intended to run headless without KVM, in such a situation of having regular keyboard doses, this seems virtually impossible.

Any ideas or remedies are thanked for.

That's all very odd...

Reply 4 2011-10-07

But I guess I'm not telling you anything! I guess at least we can rule out a kernel panic as your machine isn't completely locking up.

The dmesg log is telling us that the kernel is seeing the NIC and initialising it. So run the command again and then put it through it's paces again. Then run the command again after networking has dropped out and you've got it going again. Are the numbers (on the left) the same or do they change? The numbers are basically the time that the event happened (IIRC in seconds since boot) so if they are the same it's not anything new happening. If however they are different then the kernel has lost contact with the NIC and then reinitialised it. That might give us a clue as to what is going on.

Other than that, it might be worth checking memory usage and I/O. To check RAM and swap using the free -m -t command (-m reports MB and -t gives a totals row at the bottom). Also top is a useful commandline app which contains lots of info and updates so you may be able to see what is going on (press 'q' to quit).