Hi,

I think my NIC card stops responding every now and then.. I don't know, but this is the only conclusion I can come to w/o physical access to server.

Every few days, I won't be able to ping/tracert my server. It stops on the last hop. I have to reboot it via a remote APC port.

After it comes up, I check the logs and there is nothing weird at all. THe last thing in /var/log/messages is a server update script run as part of HSphere cron job. I'm sure this is not a problem since this script runs every few minutes and a great majority of the time, it runs fine.

I ran chkrootkit and nothing came back.

I don't believe it's an issue of server load.

Is the problem the NIC card or something else? Where can I look?

Running Redhat 7.2

Thanks

did you install any firewall scripts or snort?

What kind of nic is it? There are some crappy ones that do just crap out.

And it could also be that your server crashed at that point. (Normally you'd see weird happenings in the logs, but I've seen it happen where no significant log entries occured....)

Anyway... I'd start with the NIC and get it replaced.... make sure they put in a 3com card this time. If the problem keeps occuring... then you know it wasn't the NIC :D

I'd have someone look at console when this happens... maybe they could run 'dmesg' and see what is going on (if they are able to enter commands at all...)

If its a inksys LTNE100TX or something, and you use module "tulip"

google stuff about "tulip_old" i had a simular problem, and it was my tulip driver that was bad.

Hi,

I did faced this problem before.. the nics just all of a sudden stops working, and the system absolutely is working fine.. no errors, nothing in the logs etc.

I believe that a rc.d/network restart might solve this problem, and you can try running that every x minutes, until you change your NIC to a good known one... maybe 3com ?

:homer:

How do I find out what brand the NIC is via the shell? :blush:

Blah, it happened again... I'm just going to get the dv2 guys to replace my NIC and hope that does it. :bawling:

Originally posted by inteltechs
did you install any firewall scripts or snort?

nope.

The problem just started out of the blue... That's why I thought it was a rootkit/hacker, but I checked those out and scanners said it was nothing.

I checked netstat and there are no weird connections.

Hi,

setup a cron job for a network restart say every 30 minutes ?

If you are on redhat/mandrake

look in the file:

/etc/sysconfig/hwconf

class: NETWORK
will show you exactly what your network card is ;)


Hope this helps

:homer:

Originally posted by admin0
Hi,

setup a cron job for a network restart say every 30 minutes ?

If you are on redhat/mandrake

look int he file:

/etc/sysconfig/hwconf

class: NETWORK
will show you exactly what your network card is ;)


Hope this helps

:homer:

Thanks for the info

If I do that, it messes up HSphere because I have 15 other IPs on the NIC and HSphere has it's own script to add IPs.... maybe I should figure out how to use it :D

I did network restart before and I had to reboot it to get HSphere to add all the IPs back on because I didn't have time to figure it out back then...

class: NETWORK
bus: PCI
detached: 0
device: eth
driver: 8139too
desc: "Realtek|RTL-8139"
vendorId: 10ec
deviceId: 8139
subVendorId: a0a0
subDeviceId: 0027
pciType: 1
-

Anybody know anything about that NIC? Like does it suck :D

I've used it for over a year no problems

-
class: NETWORK
bus: PCI
detached: 0
device: eth
driver: 8139too
desc: "Realtek|RTL-8139/8139C"
vendorId: 10ec
deviceId: 8139
subVendorId: 10ec
subDeviceId: 8139
pciType: 1
-

I have the same one[as stated above]. but I have no problems with this server or it's network card.

As per my past experience [we studied this for over 2 months], I have no idea or explaination on why it does not work on 1 system and works perfectly on another system of the same config and the OS.

:homer:

Thats the problem :) Realtek NIC in a server.

Everything I use has Intel EtherExpress Pro (fxp) NIC's in them and have never experienced a problem with one of them.

Heres a quote for the realtek freebsd realtek driver:

The RealTek 8139 PCI NIC redefines the meaning of 'low end.' This is probably the worst PCI ethernet controller ever made

incase anyone else with HSphere needed to know.. this is how you restart the NIC and add all the IPS back:

[root@cp root]# service network restart;/hsphere/shared/scripts/setup-ips.pl

What type of motherboard do you have? I had this exact same problem with a server awhile ago and it turned out that that in periods of no/super low traffic the BIOS was shutting down the NIC and it wouldn't wake up on a signal.

Aaron

Originally posted by WII-Aaron
What type of motherboard do you have? I had this exact same problem with a server awhile ago and it turned out that that in periods of no/super low traffic the BIOS was shutting down the NIC and it wouldn't wake up on a signal.

Aaron

how do I tell via shell?

When you say super low, how low do you mean?

Apache averages 3 requests per second.
Current Time: Sunday, 06-Apr-2003 13:17:26 EDT
Restart Time: Sunday, 06-Apr-2003 11:29:46 EDT
Parent Server Generation: 1
Server uptime: 1 hour 47 minutes 40 seconds
Total accesses: 22309 - Total Traffic: 394.3 MB
CPU Usage: u112.86 s21.68 cu33.17 cs9.27 - 2.74% CPU load
3.45 requests/sec - 62.5 kB/second - 18.1 kB/request
13 requests currently being processed, 16 idle servers

I don't think that's the issue since it goes down all times during day and I know for fact 3 requests/sec is pretty constant on the server.

At one time I did own a couple of these 8139's and yes some of them do just stop responding. In a real production enviroment why would you been using $5 nics anyways?

Originally posted by Spingen
At one time I did own a couple of these 8139's and yes some of them do just stop responding. In a real production enviroment why would you been using $5 nics anyways?

it came with the dedicated server. :stickout:

Only top quality I see :laugh:

email them and say you would like a better one :)

I was doing more investigating and I found that cron logs stopped at the same time network connections stopped...

If the problem was the NIC, why would cron jobs stop? One would think they just keep running unless the problem is not the NIC and something else...

I have a cron job going every minute and it stops the same time the last log stops in /var/log/messages.


p.s. I got network restart going every 15 minutes, and so far so good....


:homer:

nooooooooooooooooooooooo... network restart didnt do it:


....
============================
Sun Apr 6 17:15:00 EDT 2003
Restart Done.
============================
Sun Apr 6 17:30:00 EDT 2003
Restart Done.
============================
Sun Apr 6 17:45:01 EDT 2003
Restart Done.
============================
Sun Apr 6 18:00:01 EDT 2003
Restart Done.
============================
Sun Apr 6 18:15:00 EDT 2003
Restart Done.
============================
Sun Apr 6 18:30:00 EDT 2003
Restart Done.
============================
Sun Apr 6 18:45:01 EDT 2003
Restart Done.

(it's now 19:40)

went down for an hour until i could get back to reboot it.
why do my processes just stop :bawling:

Have you asked the provider to check the console for you? There is definately something wrong, and if you are renting a dedicated server then they should look into this for year. Maybe there is something really wrong with the hardware.

Have them check console that way you can figure out if it is the whole box crapping out.

Ive seen this before with certain nic drivers...namely eepro ones. With some of the newer kernel, usually smp ones, that nic card will halt when the server starts doing some heavy processing.

My advice, compile a custom kernel or use a up kernel for now.

Wow, why not just run 'lspci'
Put down the pipe... slowly...

anything in dmesg ?

eg problems when the machine boots up?

nothing weird that I can see..
The only "negative" message I get is:
mtrr: base(0xe0000000) is not aligned on a size(0x180000) boundary

whatever that means..

Mike Harris from RedHat says that's nothing though
https://listman.redhat.com/pipermail/roswell-list/2001-September/001813.html

oh well i'm waiting for a response from dv2 now...
i'll change the NIC and go from there.