Crashing (no logs)during large file copies

egandt · n00b Joined: 24 Jul 2006 Posts: 37

I just reinstalled gentoo on a machine (the system is idential to one I used for over a year:
DFI 855 GRE (Cleron-M 1.5GHZ), Video on board
Ethernet built on Intel Gigabyte Ethernet
512megs PC3200 RAM
40GB EIDE boot drive
LSILOGIC MEGARAID S150-6 (SATA RAID5 controller)
6 Seagate 200GB drives (for a total of 1TB of RAID5 storage).
Silverstone 400 Watt PS

Changes, update the drives moved to a new case, cooler quiter.
80GB SATA boot drive
(40GB drive failed reason for the upgrade)
LSILOGIC MEGARAID S150-6 (SATA RAID5 controller), updated from BIOS 13N to 13R for drive support
4 Seagate 750GB drives (for a total of 2TB of RAID5 storage).

Problem when attempting to repopulate the RAID array I've seen 5 major hard crashes, now I see no errors in any of the logs and since the console auto blanks it-self I can not see if there was any type of kernel panic or error on screen. I've tried the Gentoo 2.6.16-R13 kernel and also a 2.6.14.7 kernel and had the problem with both. I've turned off all un-nessary services except for Samba, SSH, vix-cron, log-ng, local, net.eth0, netmount yet I still get the crash. When why or what causes it are all unknowns as I'm simple copying large quanties of data (about 800GB of data to be exact to the system), I'm about 75% of the way done now, but I can not use the storage array if this continues. The interval of the crashes is random, sometimes 6 hours sometimes 15 minutes, no patern can be deecerned. Happens with single transfers, happens with multiple transfers from 1 machine or many, happens with SMB and with FTP/SMB

Other versions: samba-3.0.22-r3, dhcpcd-2.0.5
Portage: Fully updated (07-21-06), complete install of OS and KDE
net.eth0: DHCP

MAKE.CONF:
CFLAGS="-O2 -march=pentium-m -pipe"
CHOST="i686-pc-linux-gnu"
FEATURES="parallel-fetch ccache"
CXXFLAGS="${CFLAGS}"
MAKEOPTS="-j2"
USE="X bzip2 cdr curl ethereal expat ftp gif gpm gtk gtk2 icq jpeg jpeg2k kde libg++ ncurses perl png posix sse sse2 ssl tcl tcltk tiff threads xml xpm zlib unicode pam"

1. How can I stop the console from blanking? (doing this would allow me to see if any error is printed).
2. Any ideas what the cause could be?
3. Why does dhcpcd take between 3 and 20% of the CPU when using samba?

Please help I'd hate to think I'll have to reinstalled the 200GB drives and move the 750GB drives to another system as this is suppose to be the primary backup server. (Low power low noise massive storage).

Thanks,
ERIC GANDT

Der_Faust · n00b Joined: 16 Oct 2006 Posts: 2

I have the exact same problem.

The server runs stable for weeks as long as I do not use samba. I ran memtest for a few days without errors.
As soon as I start using Samba and access some large files (> 1GB), I know the server will crash shortly. No logs, no console (keyboard does not respond), nothing... :cry:

Recently I installed a new Intel Gigabit NIC. I'm not sure but the trouble may have started then...
I will put in the old 3Com 100Mb and run some tests this week.

nixnut · Posted: Mon Oct 16, 2006 4:36 pm Post subject:

Moved from Installing Gentoo to Kernel & Hardware.
Installation seems to be done, so moved here instead.
_________________
Please add [solved] to the initial post's subject line if you feel your problem is resolved. Help answer the unanswered

talk is cheap. supply exceeds demand

Der_Faust · n00b Joined: 16 Oct 2006 Posts: 2

Well sorry it took so long. My 6 months old daughter is kind of taking up my time right now (and righteously so... :-D

)

I noticed some more events that would make the server crash dramatically:
- ftp transfers (copying an 800MB iso image would fail occasionally)
- playing music on my media box (a big playlist in winamp playing of a mapped samba share, it would stop at random)
- samba copy

Strange thing is I did a test where I copied a 10GB file a number of times (just reading though) and this would be ok all the time.
I did even more heave local file copying on the box (moving multiple 10G files around and no problem there).

Also playing the same file from a different share did not make a difference, the box still crashed. So I was pretty confident I eliminated the SATA raid controller/driver.

In the end I just swapped the old 3Com 100Mbps NIC in and the problems are gone!

Now this is obviously not what I want (bought a Intel Gb NIC + Linksys switch) since I could have stuck with the switch inside my PIX firewall...

I'm now checking for driver updated, making it modular and playing around with auto negotiation and MTU sizes.

I'll post more when there's more to be written...