02-12-2013 10:18 AM
I'm having an issue with some devices in the network showing up with the wrong port speeds in NNM. The devices are set to 1Gig interfaces, but show up as only 10Mbps in NNMi, which I believe is causing false High Utilization alarms when testing across the links.
I've put the nodes into out of service mode, and then deleted them from NNMi and then re-added them, but they still show up with only 10Mbps when they should show 1Gig.
Is there anyway I can fix this? Or any ideas?
Solved! Go to Solution.
02-12-2013 01:08 PM
Are those ports up at the time that the device is added to NNMi?
Sometimes I've seen ports reporting 10Mb, because they were down at the time of polling. Later they come up, and negotiate 100/1000, but NNMi doesn't pick that up until a config poll.
Other times the ports were actually running at 10Mb, due to problematic NIC drivers, and failing negotiation.
02-12-2013 01:30 PM
Yes the ports are always up at 1Gbps. It shows a lot of ports like this, I don't think I have that many failing ports on that many different devices. For some reason NNMi isn't polling their info correctly. Could these be a problem if the ports are auto-negotiating instead of being hard coded??
02-12-2013 01:34 PM
So I highly doubt that it's a problem with you setting the ports to auto/auto.
Out of interest, have you tried doing an snmpwalk of the device, and see if you're getting the right values reported back via SNMP?
You want to be able to eliminate potential sources of error - if SNMP reports the interface is 1Gbps, then it's an NNMi problem, and you'll probably need to log a support case. If the switch itself is reporting via SNMP that the interface is 10Mbps, well, that's a different issue.
02-13-2013 02:21 PM - edited 02-13-2013 02:21 PM
If you're not used to playing around with snmpwalk, it might be easier to start with the NNMi MIB Browser.
Go Tools -> MIB Browser.
Enter a node name.
In the OID box, enter .18.104.22.168.2.1.2. Run that query. Look at the list of results you get for ifSpeed. Note that results are in bps, so you've got lots of zeros to count on the screen.
You'll probably need to match interface IDs up against the list in the ifDescr output, to match it to an underlying interface name - e.g. Gig1/0/1.
02-14-2013 07:01 AM
Okay, so I used one of my devices and did what you said. I found that the ports that were showing up 10Mbps, indeed show 10Mbps (10000000) in the snmpwalk. These interfaces are set for 1Gig in the devices. The port numbers in the snmpwalk actually lined up with which ports NNM was saying is 10Megs.
The weird thing is, all the ports showed up as 1Gig in the snmpwalk, except the very two ports that are in the actual topology map that show up at 10Mbps. However, all port, including those two, are set for 1 Gig.
So does this mean that NNMi is reading something wrong, or my devices are sending the wrong information through SNMP?
02-14-2013 11:34 AM
I actually find that hard to believe... problem is, it's happening on 4 devices, all on the same ports, which are the only ports that are in the Topology Map in NNMi. I find it difficult to believe that Gig1/13 and Gig1/25 on all 4 devices are sending 10Mbps as their port speed when they are all set to 1Gig, and all their other interfaces are sending the information as 1Gig, and the devices they plug into at the other end of these ports all show the interfaces at 1Gig. Which obviously if a 1Gig interface was connected to a 10Mbps, the port wouldn't come up.
I find that too much of a coincidence.
02-14-2013 12:07 PM
I was also interested in your comment about the interfaces being "set to 1Gig" - do you mean they were up/up, and operating at 1Gig (as shown by show inter G1/1/1), or that they were hard-coded to 1Gig, but may or may not be up and operating?
NNMi has to believe the information that it gets via SNMP, and if the device reports the interface is 10Mbit, NNMi can't assume its anything else.
Some NMSes let you over-ride the reported speed, so that % calculations don't get screwy, but I don't think you can do that with NNMi.
02-14-2013 12:28 PM
They are Auto up at 1Gig... problem is, I just noticed that the ports have bandwidth 10000 set on them for manipulating how EIGRP chooses it's path on those links, which must be messing with the SNMP. It makes it look like the port is 10Mbps, but really, it's 1Gig. Which I believe was giving me false High Utilization alarms.
I'm retarded... I should have realized that sooner.
02-14-2013 12:32 PM
Often the mistake people make is not setting the "bandwidth" statement on Tunnel interfaces, and you get all kinds of amusing results like "Interface is 576% utilised", since Tunnel interfaces default to a very low bandwidth.