Site up/down

[Void Main]

You may have noticed this site has been up and down and when it is up it is much slower. I am having some hardware trouble with the web server. It acts like a bad network card but I have tried 3 different cards (different make/models), two different hubs and 3 different patch cables.

As soon as the card links the light flashes rapidly like it has mucho traffic, when in fact there is none. I have an old P-90 lying under the desk that was my NT4 PDC (until I replaced it with a Linux PDC) that I may migrate the web site to. We'll see how it goes. In the mean time I am going to look through the logs for anything else that might indicate where the hardware problem lies.

Here are the error messages that started a few days ago on the console and in the logs:

Code: Select all

Feb 17 05:45:30 nitro kernel: eth1: Transmit error, Tx status register 82.
Feb 17 05:45:30 nitro kernel: Probably a duplex mismatch.  See Documentation/networking/vortex.txt
Feb 17 05:45:30 nitro kernel:   Flags; bus-master 1, dirty 3377(1) current 3377(1)
Feb 17 05:45:30 nitro kernel:   Transmit list 00000000 vs. c6135240.
Feb 17 05:45:30 nitro kernel:   0: @c6135200  length 80000072 status 80010072
Feb 17 05:45:30 nitro kernel:   1: @c6135240  length 800000e2 status 000100e2
Feb 17 05:45:30 nitro kernel:   2: @c6135280  length 800000e2 status 000100e2
Feb 17 05:45:30 nitro kernel:   3: @c61352c0  length 800000f2 status 000100f2
Feb 17 05:45:30 nitro kernel:   4: @c6135300  length 800000f2 status 000100f2
Feb 17 05:45:30 nitro kernel:   5: @c6135340  length 800000f2 status 000100f2
Feb 17 05:45:30 nitro kernel:   6: @c6135380  length 800000d2 status 000100d2
Feb 17 05:45:30 nitro kernel:   7: @c61353c0  length 800000e2 status 000100e2
Feb 17 05:45:31 nitro kernel:   8: @c6135400  length 800000d2 status 000100d2
Feb 17 05:45:31 nitro kernel:   9: @c6135440  length 800000d2 status 000100d2
Feb 17 05:45:31 nitro kernel:   10: @c6135480  length 80000182 status 00010182
Feb 17 05:45:31 nitro kernel:   11: @c61354c0  length 800005ea status 000105ea
Feb 17 05:45:31 nitro kernel:   12: @c6135500  length 800005ea status 000105ea
Feb 17 05:45:31 nitro kernel:   13: @c6135540  length 800005ea status 000105ea
Feb 17 05:45:31 nitro kernel:   14: @c6135580  length 800005ea status 000105ea
Feb 17 05:45:31 nitro kernel:   15: @c61355c0  length 800005ea status 000105ea


The above messages occur on two different 3Com 3c905B cards. As I mentioned, I have swapped hardware, it happens on both 2.4.20 and 2.5.61 kernels and I've tried different 3c95x driver options (changing duplex and speed) with no improvemnet. Time to do some googlin'.

[Void Main]

Well, I think I solved that problem. I don't know why I didn't run into the problem until now. I am running 3Com 3c905B cards in two of my DMZ interfaces. The message seemed to indicate a duplex mismatch between some interfaces. I configured both 3Com cards to the same settings as my hardware firewall this handy little program and now all seems well. For some reason the autonegotiation stopped working properly. Now the network is working well again and no more error messages. Can't believe I didn't check that first thing.

[Void Main]

Ok, I think I FINALLY have pinned this network problem down. There are two or three possibilities and I can narrow it to one but I'll do that later, at least the problem appears to be fixed. The first possibility is that I have two bad 3c905B and one bad davicom card that I was using in the Void Server. The second possibility is that the 3c59x drivers in the new kernel 2.4.20 are junk and I have a 1 bad Davicom card. The third possibility is something flakey is happening with the hardware in the void server that is in turn causing the 2 3Com and 1 Davicom card to have a problem.

At any rate, the 4th card I stuck in it (D-Link) seems to be the answer to the problem. What is odd is I have been running the 2.4.20 kernel in the Void server nearly from the beginning so I don't see how that could be the problem or it would have cropped up before. That and the fact I have a 3c905B in another machine running the same 2.4.20 kernel with no problems. Oh well, maybe later I'll stick the cards in another machine and test with different kernels. I still think it's a hardware problem with that piece of junk I'm running the web site on. That machine has given me more trouble than any machine I can remember, ever. It's the last one of those "special internet deals" I'll ever buy and stick with building my own. It's a little more expensive but I get a *MUCH* better machine.

In conclusion, sorry for the trouble the last couple of days with availability (although 12 hours of that was still the fault of my provider). We'll see if we can keep the train on the tracks.