4

I am trying to configure MRTG on all of my servers for making a network weathermap, this requires running SNMP which MRTG gets the information from. So, I installed it on two servers the exact same way. However, on the 2nd server, SNMP is not able to retrieve or display the TCP information, here is the output of running
snmp -c my_community_string -v1 localhost TCP-MIB::tcp:

> snmpq TCP-MIB::tcp
TCP-MIB::tcpRtoAlgorithm.0 = INTEGER: 0
TCP-MIB::tcpRtoMin.0 = INTEGER: 0 milliseconds
TCP-MIB::tcpRtoMax.0 = INTEGER: 0 milliseconds
TCP-MIB::tcpMaxConn.0 = INTEGER: 0
TCP-MIB::tcpActiveOpens.0 = Counter32: 0
TCP-MIB::tcpPassiveOpens.0 = Counter32: 0
TCP-MIB::tcpAttemptFails.0 = Counter32: 0
TCP-MIB::tcpEstabResets.0 = Counter32: 0
TCP-MIB::tcpCurrEstab.0 = Gauge32: 0
TCP-MIB::tcpInSegs.0 = Counter32: 0
TCP-MIB::tcpOutSegs.0 = Counter32: 0
TCP-MIB::tcpRetransSegs.0 = Counter32: 0

That is clearly wrong, it also doesn't list any active connections, and MRTG also displays 0 bytes in/out (with the 3 odd spikes to about 14MB/s, I don't what might have caused that), which is impossible as I am connecting via SSH and bmon shows 2 eth0 1.11MiB 14494 26.77MiB 20011, I've made sure that the MRTG is indeed getting info from eth0, so that is not the problem. It seems that for some odd reason, SNMP is not able to compile the TCP information?

On my other server, this is a very different server, it works perfectly. This server that it works on has a differnet network NIC, hardware, datacenter and kernel but they are both running Ubuntu 13.04. The server it works on has:

*-network:0
   description: Ethernet interface
   product: NetXtreme II BCM57711 10-Gigabit PCIe
   vendor: Broadcom Corporation
   physical id: 0
   bus info: pci@0000:04:00.0
   logical name: p2p1
   version: 00
   serial: 00:10:18:ac:d9:08
   capacity: 1Gbit/s
   width: 64 bits
   clock: 33MHz

    > uname -r
    3.2.0-51-generic

The server it doesn't work on:

  *-network:0
       description: Ethernet interface
       product: I350 Gigabit Network Connection
       vendor: Intel Corporation
       physical id: 0
       bus info: pci@0000:03:00.0
       logical name: eth0
       version: 01
       serial: 00:25:90:a4:19:5a
       size: 1Gbit/s
       capacity: 1Gbit/s
       width: 32 bits
       clock: 33MHz

       > uname -r
       3.8.13-xxxx-grs-ipv6-64

Other than the TCP info, the rest of the system information from snmp seems to be okay, but I don't need any of that.

I have tried this on two servers of that spec, I got the same results on both.
Could anyone please tell me what could be causing this and if possible, how to fix it?

Goulash
  • 81
  • 10
  • That `uname -r` has an intersting output. IPv6? Is this network only using TCPv6? What does `ip addr` show on the two systems? What does `ethtool -i eth0` show on the second system? – Nils Sep 04 '13 at 20:59
  • 1
    Do versions of net-snmp differ between servers? How does snmp configuration differ between servers? What about the installed MIBs? Do they differ between servers? – Neil H Watson Sep 04 '13 at 13:54
  • Try with `sysctl -a` and check if the variables {you are interested in net.ipv4.tcp} are similar. You can also check if you have any stats at cd /proc/sys/net/ipv4 – Nikolaidis Fotis Sep 04 '13 at 12:57
  • The IPv6 flag is just nicer support for IPv6, I will try re-installing the server with the default kernel today to see if that changes anything. – Goulash Sep 07 '13 at 02:46
  • @Nikolaidis Fotis, yes the net.ipv4.tcp values are completely different. – Goulash Sep 07 '13 at 02:58
  • Oh, I have also just gotten this error when running mrtg for the first time on one of the servers that don't work: 2013-09-07 05:32:15, Rateup WARNING: /usr/bin/rateup could not read the primary log file for _eth0 2013-09-07 05:32:15, Rateup WARNING: /usr/bin/rateup The backup log file for _eth0 was invalid as well 2013-09-07 05:32:15, Rateup WARNING: /usr/bin/rateup Can't rename _eth0.log to _eth0.old updating log file – Goulash Sep 07 '13 at 03:34

0 Answers0