Just don't know what to think as now it works where it falled previously

 

Cables, cables, cables....

 

Just don't know what to think as now it works where it falled previously

 

Cables, cables, cables....

 

I have heard that one somewhere else 

I must have check them 10 times, putting them back in place to ensure they where inserted correctly. And besides that, someone else said that cables would leave some traces and wouldn't kernel panic. Why are you insisting so much with cables ? Did mine needed to warn for 2weeks to work 

It is not just a loose connector that can cause problems.  A bad cable might be one that picks up interference from nearby wiring, or induces signals where they should not exist.  Usually, this will result in some kind of data corruption. 

 

There are so many potential areas in a PC that can cause weird crashes.    One well known source is MS-Windows... 

Seriously, running memtest clears the CPU (mostly) and the memory.  It does nothing to test the disk controllers, their sensitivity to heat, or noise on the power supply lines. (The power supply may have plenty of current capacity, but how does your motherboard cope with the noise present when accessing all your disks?  )  In some cases, the one-single-rail supply might solve one problem, but cause another.  Have you tried adjusting the CPU voltage?  Is it set correctly for your CPU? Same for your memory?  They might work perfectly well when not spinning and accessing the multiple disks, but not with the additional power load.

 

I know it is tough to isolate what is the root cause.  I know that "cables" has been mentioned frequently, but you seem to be running out of tests with the existing hardware.  An intermittent error, to me, indicates a noise or voltage problem.  If you have neatly tie-wrapped the data cables from your disks to the power cables, time to cut the ties and get them further apart.  Time to up the memory voltage or cpu voltage a "tiny" bit, to see if they are more tolerant of noise... or time to drop the voltage to where they run less hot, or at a slower speed.

 

Joe L.

Finaly it didn't survived the parity build with 15 drives : kernel panic, the same as usual.

 

As this is with another power supply that can deliver 62A and just having 1 rail of 12V, the power can be cleared. So as you say memory and CPU are clear too. Anyway i will get 2x 2Go monday so i will test with some other memory. In the bios, i checked and all the timings for the memory are lower than what it can handle, so it doesn't run to its maximum speed and stress. The CPU is using its regular specifications also.

I will try to increase the memory voltage to see if it gets better, as until i get some new hardware, i can't see anymore test to do

 

So that leaves the cables, the motherboard, the 2 pci promise sata card, the graphic card, the icy dock 5 in 3 or the drives. This is going to take a month to check all this.

 

I realise when it will be done that i will have changed all the server components, except the case ;-)

 

For the wiring inside the case, the sata cables are tight together by 5 cables, indeed making a big cable for each icy dock 5 in 3. Like that the air flow is maximised. All the power cables are away from the sata cable, except of course near the icy dock where power and sata are near each other. But anyway the sata cable are said to be shield.

 

The trouble is that in France i can't find any sata cable that are know of beeing good quality and with the little thing to lock it in the sata port. I had only 4 of these shiped with my motherboard, al the others are regular sata cable, but i checked they are well inserted. If i change them, i want to use locked ones so i will be also sure it stays attached for sure.

Maybe the Promise cards and the 2 onboard "Gigabyte" SATA ports don't work together? Try disabling those 2 ports onboard but running the Promise cards still.

Either that, or the Promise cards don't work right in that motherboard - likely due to the BIOS or something on that motherboard.

 

Beyond that, start with one Promise with a single drive connected. You could even pull the SATA cages and try connecting the drives directly to eliminate them.

 

Peter

 

That is a good idea, didn't test that. I will try that tonight, thanks

Either that, or the Promise cards don't work right in that motherboard -

 

I might believe this one. on my ITX board, Sometimes it would hang during parity generation or check.

I find it just may be taxing the bus a bit.  4 drives a SATA300 on a PCI bus 

It just makes you think, When in comparison you get a 1x PCIe card that only supports 2 drives at SATA300.

 

Common Buses and their Max Bandwidth

PCI 132 MB/s

PCI 2.1 64-bit 500 MB/S

AGP 8X 2,100 MB/s

PCI Express 1x 250 [500]* MB/s

PCI Express 2x 500 [1000]* MB/s

PCI Express 4x 1000 [2000]* MB/s

PCI Express 8x 2000 [4000]* MB/s

PCI Express 16x 4000 [8000]* MB/s

PCI Express 32x 8000 [16000]* MB/s

IDE (ATA100) 100 MB/s

IDE (ATA133) 133 MB/s

SATA 150 MB/s

SATA 300 MB/s

Gigabit Ethernet 125 MB/s

IEEE1394B [Firewire]  100 MB/s

USB Full-Speed            12 Mbit/s (1.5 MB/s).

USB High-Speed (2.0) 480 Mbit/s (60 MB/s)

 

Interesting Links

http://www.pixelbeat.org/speeds.html

http://en.wikipedia.org/wiki/Peripheral_Component_Interconnect

http://en.wikipedia.org/wiki/PCI_Express

http://en.wikipedia.org/wiki/PCI-X

http://en.wikipedia.org/wiki/USB

 

 

Thanks all.

 

I know PCI is the bottleneck of my system. I go from 60Mo/s to 12 when trying with the 7 discs on the pci. I just did that because this is the only card found to work for sure with unraid. The 1500 server sold by Lime with 15 discs, uses 2 of these pci promise sata card.

 

Now if i could find someone using an adaptec sata card on PCI Express, i am interested (mini 8 ports, as i just have 1 pci expressx16 on the motherboard). I don't want to use pci express x1 as i would need too many to go up to 15 drives.

 

I tried to disable the gigabyte 2 sata ports, but same kernel panic after few hours. I tried rising the ram voltage, easy test after 10 minutes i got the kernel panic.

 

So now what you say is very interesting. The 2 promise card may have trouble with my motherboard. I just need to find people with 15 drives to see what they are using.

Dear Yodine,

 

Thank you for your efforts with you server setup.

As you might know I have the same Gigabyte motherboard, as well as 2 x TX4 PCI Promise and also 3 x ICY Dock MB455.

 

Kernel Panic never occurred again after I upgraded to the F3 bios, but I was never able to clear a disk using the 2 ports Gigabyte onboard sata controller. I mean clearing, not using them with an already cleared and formatted disks.

As clearing did not work I moved the 2 drives to the next slots, i.e. the first 2 sata slot on the first TX4. Clearing, formatting and parity built went well. I have now 8 disks with 7 TB online.

 

I now plan to just skip the onboard SATA for the moment and continue to load my system, adding new disks one by one (from my old server, doing the transition). I hope the TX4 will continue to work.

 

I am very careful because my disks are full of data.

 

MA

 

 

Thanks urmja for your feedback. I saw we where having nearly the same unraid server. We are also having nearly the same troubles.

 

But what is wierd is that they aren't the same. I could clear the drives on the gigabyte sata ports, it is just i had to do it in several times, as every operation that lasted more than 6 hours stoped by magic. So i could do 5 on the onboard sata ports (i began testing with the free version and 3 drives), then 3 by 3, except the last group where i had to do by 2.

 

The difference between us is that i started with 15 drives. I just hope you won't get trouble the day you will have more than 12 discs during the parity build. I could get mine running with 12 drives max, so i had to unplug 1 promise card for the server to build parity drive.