snmp on routed port

[night-fury]

Hi guys,

I have a routed port on our core switch and we are using that port to connect to another branch of ours through a point to point link. Here's the config. Its a 3850 switch.

interface GigabitEthernet0/13
 no switchport
 ip address 6.6.6.5 255.255.255.0
 end

I want to monitor the link bandwidth with snmp and other parameters if feasible. Is it possible to setup such traps? or any other monitoring configuration that can be done on it?

-best

nf

[Paul Chapman]

Hi NF -

You can monitor any interface on a Catalyst 3850 whether routed, switched, access, or trunk.  You just have to connect from your monitoring application with a valid SNMP community string to pull the data.  The catalyst doesn't really care from which IP you acquire the data. SNMP is a control plane protocol and is therefore bound to all IPs on the switch.

PSC

[night-fury]

Thanks for your answer Paul. I thought so but I was confused when I issued a snmp command on the interface and it didn't show up when checked the interface config.

Here's what I did

conf t

int gi0/13

snmp trap link-status

ctrl^Z

wr

and when i look at the interface config, it does not show up, i still get:

interface GigabitEthernet0/13
 no switchport
 ip address 6.6.6.5 255.255.255.0
 end

However, if on the same interface I issue: snmp ifindex persist, the config shows the same:

interface GigabitEthernet0/13
 no switchport
 ip address 6.6.6.5 255.255.255.0

snmp ifindex persist
 end

I can see the interface though, in the snmp server. Am i missing something here? please comment.

nf

[Paul Chapman]

Hi NF -

It's likely that "snmp trap link-status" and "snmp ifindex persist" are both default commands on the interface and will not show up unless you do a "show run all" command.

Based on your original question and follow up I think we need to differentiate SNMP traps from SNMP polling.  A trap is an outbound "message" from the device to a trap server which is listening for those messages.  The server may simply log the message or it may take an action, such as email the administrator.  With SNMP polling the server is responsible for "asking" the device for information.

From the switch's perspective, things that are event oriented, like Interface-Up and Interface-Down are also designated as trap type events.  Things that are information oriented, like port statistics, can only be acquired by SNMP polling.

Circling back to the 2 interface commands you pointed out... Some Cisco platforms (particularly Catalyst 6500, which could have more than 500 interfaces) have those traps disabled by default.  These platforms tend to get a lot of port state changes that are uninteresting to administrators and eat precious CPU cycles to generate.

For your 3850, it's unlikely that you will need any interface level commands to get your SNMP set up.  All the commands should be part of the global configuration.

HTH

PSC

[night-fury]

Thanks Paul.

This is really an informative reply. Appreciate it. it assures I have everything in place from a switch point of view. I will check from snmp server end if any tweaking is required. And yes, i did see the link-status snmp entry when did a show run all.

-best

nf