> Hi, > > After deploying some new Dell PowerEdge 1950 running RHEL 5.2, I've > seen the following problem : The cacti/snmp graphs for a bonded > interface made from the two ports of a dual PCIe Intel NIC were wrong. > > After investigating, here's what I currently see : > - The values ifconfig reports seem wrong too > - The values found in /proc/net/dev seem wrong too > - Using "iptraf", I don't see the bonded interface, but the values for > the two physical interfaces seem good (!?) > > Values : > - From snmp : 800Mbps for both eth2 and eth3, 1.6Gbps for bond1 > - From ifconfig and /proc/net/dev : About 100MB/s for the eth > and 200MB/s for bond1 (rough estimate, but matches what's seen with > snmp) > - From iptraf : 190Mbps for both eth2 and eth3, which means a total of > 380Mbps, which is the much more realistic value and about 1/4th of the > value reported by snmp/ifconfig. > > The system has been installed as 32bit and is running the > 2.6.18-92.el5PAE kernel, as x86_64 was not an option here. The bonded > interfaces are using the e1000 kernel module. > > Does this problem ring a bell to anyone? I don't recall seeing it > before, but I don't think I've ever had a 100% identical setup, which > could explain.
I have found iptraf to be unreliable for network bandwidth usage measurements beyond ~100mbit/s. Measurements are done from the switches instead in our environment. A lot less painful and much more accurate. -- Jussi _______________________________________________ rhelv5-list mailing list [email protected] https://www.redhat.com/mailman/listinfo/rhelv5-list
