My little server freezing

pie86 · 2009-10-28 09:43:14

Hi everybody,
I have a problem with my brand new fitpc2 working as a little file server at home. It works like a charm for one hour (min) to 16h (max) with transmissiond lighttpd and bftpd, then it freezes: the keyboard does not respond (no maiusclock and numlock light appears if i press the corresponding keys), the ethernet card is active and the green light is flashing but I cannot acquire any packet with wireshark and the power button (which I use to shutdown the pc with an acpi rule) doesnt shutdown anything (I have no screen so I cannot see anything from ttys). According to me it isn't a kernel panic becouse numlock and maiuscblock aren't flashing (on my usb keyboard).

Logs doesn't report anything bad except for those @ in messages.log

Messages.log

Oct 28 04:44:20 argo -- MARK --
Oct 28 05:04:20 argo -- MARK --
Oct 28 05:24:20 argo -- MARK --
Oct 28 05:44:20 argo -- MARK --
Oct 28 06:04:20 argo -- MARK --
Oct 28 06:24:20 argo -- MARK --
^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@^@Oct 28 09:54:37 argo kernel: Linux version 2.6.31-ARCH (root@architect) (gcc version 4.4.1 (GCC) ) #1 SMP PREEMPT Fri Oct 23 11:12:58 CEST 2009
Oct 28 09:54:37 argo kernel: KERNEL supported cpus:
Oct 28 09:54:37 argo kernel: Intel GenuineIntel
Oct 28 09:54:37 argo kernel: AMD AuthenticAMD
Oct 28 09:54:37 argo kernel: NSC Geode by NSC

Errors.log is empty

Daemons.log

Oct 26 18:52:51 argo init: Entering runlevel: 3
Oct 26 18:52:58 argo acpid: starting up
Oct 26 18:52:58 argo acpid: 2 rules loaded
Oct 26 18:52:58 argo acpid: waiting for events: event logging is off
Oct 27 08:14:36 argo init: Entering runlevel: 3
Oct 27 08:14:43 argo acpid: starting up
Oct 27 08:14:43 argo acpid: 2 rules loaded
Oct 27 08:14:43 argo acpid: waiting for events: event logging is off
Oct 28 09:54:36 argo init: Entering runlevel: 3
Oct 28 09:54:43 argo acpid: starting up
Oct 28 09:54:43 argo acpid: 2 rules loaded
Oct 28 09:54:43 argo acpid: waiting for events: event logging is off

syslog.d

Oct 26 18:52:52 argo syslog-ng[1018]: syslog-ng starting up; version='3.0.4'
Oct 27 08:14:37 argo syslog-ng[1036]: syslog-ng starting up; version='3.0.4'
Oct 28 09:54:37 argo syslog-ng[1047]: syslog-ng starting up; version='3.0.4'

kernel.log

Oct 28 01:18:42 argo kernel: Machine check events logged
Oct 28 03:04:19 argo kernel: Machine check events logged
Oct 28 09:54:37 argo kernel: Linux version 2.6.31-ARCH (root@architect) (gcc version 4.4.1 (GCC) ) #1 SMP PREEMPT Fri Oct 23 11:12:58 CEST 2009
Oct 28 09:54:37 argo kernel: KERNEL supported cpus:

Smartctl says hd has no problems and for the moment I can't run a memtest becouse I have no screen.

Any ideas?
thanks,
pietro

eldragon · 2009-10-28 12:13:32

hmm, im not sure about this, but does a kernel panic reveal itself to a usb keyboard?

ive had a similar problem a couple of weeks ago with an old nf4 server, turned out to be a faulty ram (i did have kernel panic flashes in my ps2 keyboard though). the thing could stay up for about a day...if you have several sticks of ram, remove some...

what about the PSU? are the power rails within acceptance?

and last but not least. thermal reports? check /sys/class/thermal

seiichiro0185 · 2009-10-28 14:46:01

Hi,

maybe try the sync, umount and reboot shortcuts from the kernel-section here. and see if anything in the logs after this. (I suspect you are doing a hard shutdown since nothing is reacting anymore?)

pie86 · 2009-10-28 18:27:03

seiichiro0185 wrote:

(I suspect you are doing a hard shutdown since nothing is reacting anymore?)

Yes, you are right, at least 10 hard shutdown until today!
I'll try to reboot with the key combinations next time it freezes.

@eldragon: unfortunately inside it looks exactly like this:
http://www.fit-pc.info/english/images/f … -bot-m.jpg
http://www.fit-pc.info/english/images/f … -top-m.jpg

so I don't think there can be a way to remove ram :S

How can I test the PSU?

According to me (and lm_sensors) there are no sensors on the pc! In /proc/acpi all temperatures are 0.

eldragon · 2009-10-28 18:37:15

i thought we were talking about a full computer. set as a media server..

i guess the best shot is to do a lspci -vvnn and search the web for related issues with your hardware and linux. build a custom kernel with debugging enabled and keep hunting for errors. thats all i can think of right now.

that kind of hardware is not likely to overheat (no mounting holes for a cooler) if unsure, tap it with your finger, if you get a blister, then its definately heat related RMA the unit back.

pie86 · 2009-10-29 11:53:09

One good news, keyboard shourtcuts works!

Now I'm compiling the kernel in debug mode...let's pray!:)

pie86 · 2009-10-30 10:47:59

Compiling the kernel is making me loose more time then I expected...

However, today I was lucky because I could see the freezing process just while it was happening, and it's making my wonder about an ethernet driver problem.
Here's why:
I was transfering the kernel compiled on my pc (127mb) via ftp, at the same time I was using ssh and downlaoding from bt. Suddenly I've lost ssh connection and ftp transfer. After 1sec ethernet's card green light start flashing random and router ignore it.

Exactly the same thing happened while transfering the same file via ftp just a few minutes later.

If I'm right there are bad news for me: https://bugzilla.redhat.com/show_bug.cgi?id=252955

When I'll finish installing the new kernel (where is included rt3090 driver needed for the wireless kard) I'll try to use wireless card and see what happens.

thank you everybody
Pietro

pie86 · 2009-10-31 19:00:44

I was right, there's an error in the b8169 module in 2.6.31. Installing the driver from realteck (this patch is needed: http://www.jamesonwilliams.com/hardy-r8 … ment-1036) and blacklisting b8169 (b8168 has to be loaded) stops the system freezes.

I read somewhere in the web that kernel >= 2.6.30 solved this issue but probably they were wrong (or referring to another problem).

I haven't tried last kernel (2.6.32) yet.

I was wondering about adding a page about fitpc on the wiki? what do you think about it?

thanks everybody,
Pietro

Arch Linux

#1 2009-10-28 09:43:14

My little server freezing

#2 2009-10-28 12:13:32

Re: My little server freezing

#3 2009-10-28 14:46:01

Re: My little server freezing

#4 2009-10-28 18:27:03

Re: My little server freezing

#5 2009-10-28 18:37:15

Re: My little server freezing

#6 2009-10-29 11:53:09

Re: My little server freezing

#7 2009-10-30 10:47:59

Re: My little server freezing

#8 2009-10-31 19:00:44

Re: My little server freezing

Board footer