strange network problem...
 

i cant track this one down.

my wife takes online classes and is required to spend time reading and posting to message boards provided by the school (regis university, if that helps). the problem is that every time she connects to one of the message boards, or cable connection pukes. its might be right away, it might be a few minutes, but it always dies. the modem itself appears to still be connected (connection light), but the only way to fix it is to cycle power to the modem while rebooting the system, the system refuses to release the ip via software after this happens. the problem happens in all the following configurations:

windowsxp, internet explorer
winxp, mozilla (1.2.1?)
red hat 8.0, konqueror
red hat 8.0 mozilla (1.1)

of note is that xp and rh8 are on different hard drives and the xp drive is dying (im replacing it tonight). it occurs to me that i never bothered to look at the lights on the nic when this happens, so i might have missed something important. we have emailed the webmaster of the site and he has never seen the problem. i havent called the cable provider because theyre fucking clueless.

im tempted to pick up a nic (just a cheapo 3com) to see what happens, but im not sure its worth the money. any ideas?

~james

I'll save you some time, if you call the cable provider, here's what they'll say: reboot, reload, reinstall.

Because this is a thing that should never happen - nothing in the protocol could possibly cause it, so there is something else at work and you'll have to do more troubleshooting.

Need more detail. How do you know it's the connection that's locked up? Do you open other apps that use the net and find they don't respond?

Do you have a software firewall? Do you connect through a proxy server?

Try turning off java, javascript etc.?

after shes been using the forum for a short while, she will attempt to open a page and get a page not available message appropriate to the browser shes using. at that point, no page will open and i am unable to ping, ftp or release the ip. system resources still look okay, but there is no network connectivity whatsoever.
zonealarm under windows, the default medium firewall settings at install tiime in redhat. it didnt occur to me to try disabling zonealarm before my xp hard drive died, and im not smart enough to have figured out how to disable the linux firewall.
nope.
damn, i forgot to mention that. under windows, if we disable all scripting in mozilla only (this doesnt help in ie), it works for significantly (maybe an hour) longer before puking (this is why im so hesitant to try a new nic). same thing under redhat doesnt help.

~james

Is this forum browseable by the rest of us, or is it closed? (Got an URL?)

What's the forum software?

its closed, unfortunately. i cant determine what the software is (its not marked, may be custom) but its (i think) php. it works just fine from every other computer ive tried.

~james

You mentioned buying a NIC... how is the system connected to the cable modem now?

direct. motorola modem connected directly to the nic, which is an intel vanilla 10/100. the modem is also usb capable.

~james

And has this *ever* happened using any other site?

come to think of it, when i got on the computer this morning the connection was dead (in linux) but i rebooted the system (didnt touch the modem) and it was fine. now that i think about it, that does happen 1-2 times a week. ive had that problem though on different computers using different cable providers, so i always figured it was something to live with.

but to answer your question, no, it never happens on any other page ive opened. i have gone out looking for pages that might cause problems, overuse of javascript, etc. but havent found anything that could do what that forum does.

~james

Could be some problem between interrupts on your system that a new NIC card might address by doing plug-n-play better.

It may be something really wacky - like, it's not the software that the forum is running, but their implementation of TCP/IP may exercise some partly-broken bit of the protocol on your hardware. So you hit it once a week and figure it's just life, but the wife is dealing with a connection that every once in a while hits a rare part of the protocol that just doesn't work on your setup. Maybe something that's reserved and which is ignored on most setups but causes the card to go into a wait state on yours. Maybe a combination of commands that every other modem handles OK but yours doesn't. Who knows!

The tech support thickies will still insist that you reinstall or reload, because 99.99% of the time, such problems are driver problems. So your idea to change out the NIC sounds like exactly the right way to go. And change out the cable just to rule it out. And if that doesn't do it, call tech support and convince them that they need to change out your modem, if you have a lease deal or such.

The nice thing is that you don't need to try an expensive NIC, just a $20 deal. But I'd try one with a different chipset than the one you have. When I last looked into these things, Intel had a different chip set than everybody else, but I don't know if that's still true.

sounds good. ive reinstalled both redhat and xp out of frustration dealing with this problem, so im at the end of my rope. im gonna pick up a 3com tonight (ill doublecheck chipsets, but i think youre right about intel) and see how it goes. thanks for the advice!

~james

Do I understand your post correctly? When the connection to school crashes, then no other task works? You cannot even Cntrl-Alt-Del to load and execute Task Manager?

If so, failure information may exist in system logs. I assume the obvious - Device Manager reports no hardware conflicts.

Very little hardware can cause a whole OS lockup. Just wondering if a task is consuming so much CPU time - a problem with programs that execute CPU intensive Java scripts - as to make other tasks appear locked. That and other symptoms would be obtained from Task Manager.

Naw, t, his network connection dies and resetting the cable modem doesn't help; he has to reboot completely in order to get networking back.

Strangely, that's actually my ISP's support script for their cable modem subscribers....

And yes, I've had to do it...i have no idea why it works, but it did.

Dagney

it appears to be working with a supercheap (5 bucks after rebate!) netgear nic. i have not yet installed zonealarm, so we will see what happens there. but i have my fingers crossed.

tw, no it was just internet access, other apps worked fine as long as internet access wasnt needed.

~james