6

On the SG300-52, I've noticed that SNMP monitoring of LAGs versus regular switch ports give incorrect (way too large and unrealistic) values for the octets transferred.

For simplicities sake, let's say I have two devices connected a SG300-52 switch. One of these devices is using LAG (802.3ad) with GigE ports while one is not (vanilla GigE). Transferring data from device 2 (non-LAG) to device 1 (LAG) shows an average rate of ~80MB/s on device 2 and device 1 using software interface monitoring tools. Monitoring SNMP ifOutOctets the LAG port for device 1 shows an average bandwidth (so the derivative of the raw SNMP value) of ~320MB/s. Looking at ifInOctets for the non-LAG device 2 port shows an average bandwidth of ~80MB/s.

Even ignoring the implausibly large value for bandwidth of the LAG device, communication to this device is only from the non-LAG device, i.e. there aren't other devices on this switch that are communicating with the LAG device at the same time as this test.

Using ifHC*Octets yields even more ridiculous results for the LAG group, while non-LAG ports still report bandwidth correctly. Has anyone else come across this?

Improve this question
6
  • 1
    Why is the value implausibly large? How many ports are in your LAG? Are you sure you're computing the rates correctly?
    – Sebastian Wiesinger
    Commented Jun 6, 2015 at 11:33
  • Just 2x 1-GbE ports in the LAG. The values returned for instantaneous BW (computed as the time derivative of ifOctets* by InfluxDB) are ~320 MB/s for a single direction. Excluding framing, padding, scrambling and other variables, this would suggest that the aggregate link bandwidth is in excess of 24Gbps.
    – rub-a-dub-dub
    Commented Jun 7, 2015 at 5:40
  • 1
    320MB/s would be 2.5 GBit/s. That is not so inplausible if you would've had more than 2 members and there would perhaps be other traffic that could account for these number. As I don't know InfluxDB I can't say anything about the numbers. If it is correct it sounds like a counter bug. See my answer for that.
    – Sebastian Wiesinger
    Commented Jun 7, 2015 at 20:21
  • @Sebastian: Whoops, that's right 320MB/s ~ 2.5 Gbits/s. Moot point anyways since I only have two members in the LAG group, so maximum line rate in a single direction should be 2 Gbits/s.
    – rub-a-dub-dub
    Commented Jun 11, 2015 at 1:21
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you could provide and accept your own answer.
    – Ron Maupin ♦
    Commented Aug 11, 2017 at 16:40

1 Answer 1

Reset to default
1

The thing is, Cisco is kind of notorious for wrong counters, be it interfaces or other stuff, especially "virtual" interfaces like LAGs. Maybe you're hitting a bug. You could try another IOS version or look for a Cisco bug report for this. You could also open a TAC case to have them look at it.

Also try looking at the counters for the physical member interfaces of the LAG. See if they in sum are correct / divert from the LAG counter.

Improve this answer
1
  • 1
    SG300 series switches do not run IOS.
    – YLearn
    Commented Nov 4, 2015 at 23:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.