[Trilug-ontopic] ifconfig question.
Don Jerman
djerman at pobox.com
Sun Feb 3 14:12:26 EST 2008
There you pass me. A quick grep or two shows that they're unsigned
longs from /proc, so they'll roll over every 4 gigabytes or 18
exabytes, depending on what %ul means to your architecture. They
probably overflow independantly but I'm not able to delve into the
/proc sources so quickly, and it may be driver-dependent.
The best way to factor it out is to snapshot the statistics frequently
enough to get a rate and not have to worry about overflow. Every 5
minutes would do for a start unless you're on a rather busy network.
If you'd like to roll your own snapshot from the /proc output, the
parsing for ifconfig is done in the lib/interface.c file in the
net-tools source tree - you could get the field order from there.
Besides, if this isn't where your problem is, then it probably doesn't
matter :).
On 1/29/08, Jason Watts <jsnonzzr at gmail.com> wrote:
> Don,
>
> Thanks for the answers.
>
> After chasing this down some more... it looks like the interface that I
> coppied that from isnt really were our problem is (though we are planning on
> changing the duplex stuff anyways.)
>
> When the counter overflows and resets, does just the one counter reset, or
> will it reset all of them, including the overrun counter? (the box has
> acutally been up for 31 or so days.)
>
>
> On Jan 29, 2008 8:06 PM, Don Jerman <djerman at pobox.com> wrote:
> > The questions are quick...
> >
> > Looking at this I'd guess uptime has been a few hours, or else several
> > months, because the counters reset when they overflow or when you
> > restart the interface.
> >
> > Error rate is around 0.027%, so it might not be worth fixing. Your call.
> >
> > Overruns are typically caused by the next packet arriving before the
> > interrupt service routine is done emptying the buffers. That would
> > indicate a fast sender at the other end, many senders talking at once,
> > or a slow driver or malfunctioning card. Check for duplex mismatch
> > too. That's about all I can infer with this information (not that I
> > know all that much more about it).
> >
> >
> >
> >
> > On 1/29/08, Jason Watts <jsnonzzr at gmail.com> wrote:
> > > Guys, hopefully these are two quick questions.
> > >
> > > when I do an ifconfig -a on one of my boxes, I see a fairly large number
> of
> > > overruns.
> > >
> > > UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
> > > RX packets:173319141 errors:0 dropped:0 overruns:46678 frame:0
> > > TX packets:194764057 errors:0 dropped:0 overruns:1 carrier:0
> > > collisions:0 txqueuelen:100
> > > RX bytes:1809166178 (1725.3 Mb) TX bytes:102562715 (97.8 Mb)
> > > Interrupt:11 Base address:0xb800 Memory:ec0c3000-ec0c3038
> > >
> > > My 2 questions.
> > >
> > > 1) How often do these numbers reset?
> > >
> > > 2) to me, 47k seems like a large number, especially when all my other
> boxes
> > > are showing none, but what would be an acceptable number? (I realize
> this
> > > is relative, but I'm asking for opinions as this is the first time I
> have
> > > ever researched this stuff and have little idea.)
> > >
> > > Thanks for any input and advice.
> > >
> > > jsn
> > >
> > > _______________________________________________
> > > Trilug-ontopic mailing list
> > > Trilug-ontopic at trilug.org
> > > http://www.trilug.org/mailman/listinfo/trilug-ontopic
> > >
> > >
> > _______________________________________________
> > Trilug-ontopic mailing list
> > Trilug-ontopic at trilug.org
> > http://www.trilug.org/mailman/listinfo/trilug-ontopic
> >
>
>
> _______________________________________________
> Trilug-ontopic mailing list
> Trilug-ontopic at trilug.org
> http://www.trilug.org/mailman/listinfo/trilug-ontopic
>
>
More information about the Trilug-ontopic
mailing list