After reading all the posts on the Intel cards and eepro problems, I thought I
would just mention that over the past 2 weeks, I have noticed that my NIC cards
start to exhibit the following symptoms: When pinging to router on the local
network, ping times will go from < 1ms and start climbing to 1000ms or more.
Sometimes it gets to the point where I cannot access the server at all,
remotely or via console. I get a message stating that an attempt to restart the
transmitter has failed. What I have done is put a bandaid on the situation and
cron restarts the network daemon every 10 minutes. Not what I want to do and
sometimes even this is not enough. I am running a RedHat 6.0 box with 2.2.5-22
kernel. One of my other servers started to exhibit this same problem about the
same time. It is also RH 6.0 with 2.2.5-22 kernel. I originally had this
problem on both servers when I first installed them (2.2.5-15 kernel) and a
kernel upgrade fixed it. But now it is back. With all the discussion, has
there been a consensus on any fix for this problem? I am running Intel cards
with the eepro100 driver dated Jun 2 1999.
Thanks
Chris
Brian wrote:
> Not to answer his questions for him, but his posts on June 8th (I asked a
> similar question) suggest this is a remote system with one NIC. Without some
> other access method, there's no way to tell the difference between a NIC and
> system hang.
>
> Compounding the problem, probably some or all of the syslog file gets lost
> when the power is cycled. It's very frustrating to find the current boot log
> as the first item in syslog.
>
> So Scott... if possible, get a second, non-intel card in that server and
> setup a static route between the server and some other box using that second
> card. If only the NIC hangs, you can come in that static route to ifup/ifdn
> the card and check out the logs.
>
> Meanwhile, I'm going to try 1.20.2.10.
> On my system, it took the older drivers 1.5 days to hang, so I'll let you
> know Tuesday.
>
> -- Brian
>
> On Mon, 12 Jun 2000, Andrey Savochkin wrote:
> > Hello Scott,
> >
> > On Mon, Jun 12, 2000 at 12:07:28AM -0400, Scott R. Every wrote:
> > > Well, the reason I think its a network problem, is that this machine has
> > > never had any trouble running heavy load w/o heavy network traffic. as
> > > soon as the traffic starts up, the network goes away.
> >
> > You haven't explicitly answered my question.
> > Does the whole system hang or just network stop?
> > Do you use the word "crash" in its literal meaning or it's just a
> > hyperbole?
> >
> > > there are NO messages of any sort in ANY log. If anyone has suggestions
> > > for how to better log this, i'd love to hear them. at this point i have
> > > a
> >
> > If there had been messages they should be captured with your syslog config.
> > You may explicitly redirect all kernel messages to a file, like
> > kern.* /var/log/kern
> >
> > Best regards
> > Andrey V.
> > Savochkin
> >
> > > production machine that crashes daily. even NT works better than that...
> > >
> > > s
> > >
> > > --On Saturday, June 10, 2000 5:54 PM +0800 Andrey Savochkin
> > >
> > > <[EMAIL PROTECTED]> wrote:
> > > > Hello Scott,
> > > >
> > > > On Fri, Jun 09, 2000 at 05:46:47PM -0400, Scott R. Every wrote:
> > > >> just an FYI.
> > > >>
> > > >> machine died again today. heavy network load. sending out 2 large
> > > >> mail lists, with lots of web traffic and a few quake servers running.
> > > >> this is using the 1.20.2.10, which is the latest AFAIK. No messages
> > > >> like the ones Adrey sent below appeared in the logs. Did I mention
> > > >> this is an SMP server? Would that make a difference?
> > > >
> > > > I've just noted the word "died".
> > > > Do you mean that the whole system hanged, not just network
> > > > communications?
> -
> To unsubscribe from this list: send the line "unsubscribe linux-net" in
> the body of a message to [EMAIL PROTECTED]
--
Chris West - N5LTC
Sys Admin/Comm Tech
WCSOnline/Western Communications
915-949-3000
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
