ifconfig byte count limit?

slylittlefox · 6 Sep 2009 at 01:55

I've just found something curious - ifconfig seems to go all the way around the clock when measuring transfers:

Code:

lm@box:/root$ ifconfig
eth0      Link encap:Ethernet  HWaddr 00:11:11:82:9e:8c
          inet addr:192.168.69.10  Bcast:192.168.69.255  Mask:255.255.255.0
          inet6 addr: fe80::211:11ff:fe82:9e8c/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:509501903 errors:0 dropped:0 overruns:0 frame:0
          TX packets:489548401 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          [B]RX bytes:3277153981 (3.2 GB)  TX bytes:1777891207 (1.7 GB)[/B]
          Interrupt:16

Then no more than 10 mins later:

Code:

root@box:~# ifconfig
eth0      Link encap:Ethernet  HWaddr 00:11:11:82:9e:8c
          inet addr:192.168.69.10  Bcast:192.168.69.255  Mask:255.255.255.0
          inet6 addr: fe80::211:11ff:fe82:9e8c/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:510523358 errors:0 dropped:0 overruns:0 frame:0
          TX packets:490088329 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
         [B] RX bytes:524151589 (524.1 MB)  TX bytes:1821727108 (1.8 GB)[/B]
          Interrupt:16

What the hells going on?

slylittlefox · 6 Sep 2009 at 01:59

Ah, should have 'Let me Google that for you'

The counters have a hard limit of 4GB - mystery (albeit a very small, quick one) over

Next question - is there a way to keep a continuous count?

tntcoder · 6 Sep 2009 at 02:17

Are you on x86?

My x64 box does > 4gb.

Code:

[jack@tmain ~]$ ifconfig eth0
          RX bytes:8881205138 (8469.7 Mb)  TX bytes:316939370 (302.2 Mb)

At a guess I would say it could be architecture related, with a 2^32 limit on x86 systems. (Could be very wrong though

)

If you want a permanent counter, I would make a script to extract the value to a file or a bash variable and do the maths when it resets, and cron job it for once a minute.

slylittlefox · 6 Sep 2009 at 02:47

tntcoder said:
Are you on x86?

My x64 box does > 4gb.

Code:

[jack@tmain ~]$ ifconfig eth0 RX bytes:8881205138 (8469.7 Mb) TX bytes:316939370 (302.2 Mb)

At a guess I would say it could be architecture related, with a 2^32 limit on x86 systems. (Could be very wrong though )

If you want a permanent counter, I would make a script to extract the value to a file or a bash variable and do the maths when it resets, and cron job it for once a minute.

Ah, I did think it might be related to that. It seems a bit silly though, why would it not just store the value as a string and update that? That is, the string/int 4001MB is not 4GB in size

Also gave some thought to a script, but it's too late for me to work the logic out. Here's where I'm at in pseudocode-BASH:

Code:

!/bin/bash (haven't got a hash key - Mac :o)
totalcount=`cat /opt/ifcount/totalcount`
lastcount=`cat /opt/ifcount/lastcount`
count=`ifconfig | awk {operation}`
if [ count < lastcount ]
then
    newcount = $totalcount + $lastcount
    echo $newcount > /opt/ifcount/totalcount
    echo $count > /opt/ifcount/lastcount
else
    echo $count > /opt/ifcount/lastcount
fi

Thing is, I'd have to run it every 10s or similar, otherwise potentially a lot of data could get 'lost', depends on what the performance impact was of running such a small shell script.

Thanks for the reply

GarethDW · 6 Sep 2009 at 03:31

If you know that the value is being reset at 4GB then why not keep a count of the number of times that it's been reset?

Similar to what you're doing above, query the ifconfig count and compare it to the last value, if it's lower then you can assume the 4GB limit was reached and can increment the reset count by one. To tally up the total, just multiply the reset count by 4GB and add on the current ifconfig count.

Shouldn't think it would need to be run that frequently.

Alternatively, there are plenty of tools you can use to monitor bandwidth, such as bandwidthd or Cacti.