Hi,

I'm having an odd issue on a new server I built. Every couple days I notice my server is off. I turn the server back on and check the logs and there is nothing in the logs that indicate any sort of problem. Maybe syslogd is not configured to log the right events but there is no indication of any sort of problem. The machine just randomly turns itself off. I'm unsure if it is a clean shutdown or not, I'm assuming it isn't because I believe that should appear in the logs. I monitor the temperature, memory use, and load on this machine and dump it to some graphing software. There is no jump in temperature, load, or memory use prior to the crash. The majority of times this occurs the machine is completely idle. I have ran memtest86+ and it has no errors. I have only witnessed a shutdown once and it appeared that X quit followed by the machine turned off. The machine was idle when this happened.

The hardware configuration is the following:

Motherboard:
ECS NFORCE 570 SLIT-A (V5.1) LGA 775 NVIDIA nForce 570 SLI ATX Intel Motherboard

Processor:
Intel Pentium D 805 Smithfield 2.66GHz 2 x 1MB L2 Cache LGA 775 Processor

Video Card:
MSI NX7100GS-TD256E GeForce 7100GS 256MB 64-bit GDDR2 PCI Express x16 Low Profile Video Card

Ram:
Patriot Signature Series 512MB 240-Pin DDR2 SDRAM DDR2 667 (PC2 5300) System Memory
Kingston 1GB 240-Pin DDR2 SDRAM DDR2 667 (PC2 5300) Desktop Memory

Hard Drives:
2 120 GB IDE drives
1 27 GB IDE drive

PCI cards:
Hauupauge PVR500
Hauupauge PVR350

This machine is running fedora core 6 32 bit.

There is only 1 unresolved configuration problem that I have that might be related. The NFORCE motherboards have a bug in their bios that cause the ACPI timings to be incorrect. To fix this i've enabled the kernel flag 'acpi_skip_timer_override'. Which should fix this bug. I have not tried disabling ACPI entirely.

Anyone have any hints on what I might do in order to track down what is causing these random shutdowns or any idea on what is causing them?

Thanks

The first thing that hit me is the odd memory combo.

I always tell people if you're going to have two sticks of memory, make sure they are identical, even if the mobo manufacturer says it is not imperative. DDR2 is dual channel memory meaning you have two white slots and two black slots (or two purple & two orange in this case), the two sticks should be both in the orange slots and "equal". Meaning "same make & model & size".
I highlighted your mobo specs & stuck them into the google search field and this is the first link I looked at, look at what people say about memory usage.

http://www.newegg.com/Product/Produc...82E16813135026

Also, this is an economy class mobo, you get what you pay for. It does not supply enough power for the processor it supports (shy of one 4 pin power connector). And there may be a heat issue at hand here if there is not enough cooling (Case fans), as there is only one 3 pin fan connector, this type of processor should have at least two case fans in a relatively cool setting, and a good quality heatsink and fan on the processor.

It is defiantly not a heat issue. I have a zalman 9500 on it and the cpu stays around 25 degrees. 35 under load. I have 3 case fans also. Power may be an issue but it is odd that it is occurring when idle. Also I agree having 2 different memory sticks is not ideal but memtest86+ ran through twice with no errors. I have added an entry in syslog.conf that logs everything to 1 file. Hopefully the next time it shuts down i'll have something in that log. If I don't get anything in the log I'll probably test it with just the 1 GB ram stick.

It'l be slower with only one channel of memory, the memory issue should not be responsible for the shutdowns, only a performance penalty. But then again one person in that article I looked at mentioned how it would not run with the wrong memory, I have no experience with that board per say but if it runs so far, you can pretty much say that does not apply here. I would have first looked at temperature as modern boards will shutdown when there is a problem in that area, but it looks like you got good air movement, hopefully it is not in a cabinet as I've seen that being the culprit for some people where just removing the door off the cabinet stopped the random shutdowns.
Next item I would look at is the power supply box, if you are around when it happens again, wrap a hand around it to see how hot it is. You more than likely have it plugged into a UPS or surge protector at least. If you don't have it in a good UPS, or even if you do, plug a digital clock radio in the same source to see if it blinks after the computer shuts down, indicating an interruption.
I would not think it is OS related but I would not rule it out yet. Not sure about FC6's power saving settings as I've not been there yet, you probably had a peek at that I'm sure. And bios settings in that area also. I tend to lean more towards hardware on custom built machines as there are allot of different hardware that don't get along and extensive reading is essential before purchasing the parts. I do allot of computer repair and find when "odd" things happen, it is more likely to be with the custom jobbers, or a dust build-up in the box. It is not uncommon for certain software packages settings and/or hardware settings to automatically shut down after a period of in-activity also, the hardware part is usually set through the bios. In the bios under PC Health status is where one can set the auto shutdown temperature. Under Power Management you can set it to re-boot after such an event automatically, but you probably would rather figure out why it shuts down first.

Today I witnessed another outage but this one was caused by a power flicker. After inspecting my UPS my battery is either completely dead or the UPS is hosed. On Monday i'm going to go to a local battery store and see if they can replace the battery. I'm hoping that this is simpily what the problem was. Though if it is I should probably be concerned with the power at my house flickering every couple days