Solved: 2960 does not respond to SNMP queries

cweatherford · ‎09-22-2010

I have SNMP set as follows:

snmp-server community xxxxxx RO
snmp-server community xxxxx RW

But using SolarWinds NCM I can poll half of my 400 switches just fine but the other half do not respond to SNMP. I would apprecaite any thoughts!!

Configs SHOULD all be the same except for host, crypto and IP addresses.

Thanks!

Kostas Kyriakos · ‎09-26-2010

Hello.

Pretty sure you need snmp-server host configured. Try that and then add snmp-server enable traps chassis for example and see if that gets you anywhere. Then try pinging the SNMP_SERVER_IP from inside one of the non-responsive switches. If it fails then it's most likely a routing problem. Try adding snmp-server source-interface traps and snmp-server trap-source . And you can always try debug snmp sessions and/or packets.

Hope this helps.

View solution in original post

Kostas Kyriakos · ‎09-26-2010

Hello.

Pretty sure you need snmp-server host configured. Try that and then add snmp-server enable traps chassis for example and see if that gets you anywhere. Then try pinging the SNMP_SERVER_IP from inside one of the non-responsive switches. If it fails then it's most likely a routing problem. Try adding snmp-server source-interface traps and snmp-server trap-source . And you can always try debug snmp sessions and/or packets.

Hope this helps.

jonathanaxford · ‎09-26-2010

Hi there,

Having the snmp-server community STRING RO/RW command should be sufficient to get ORION to poll the switch.

As KF K says, try some test PINGS to make sure that the devices in question are accessible from the ORION server and vice-versa, these sort of problems are usually down to connectivity. One more thing to check is that the community string value is in the same case on both the server and switch.

I think that the snmp-server host command tells the switch where to send SNMP notifications/traps to, rather than restricting SNMP polling to a specific host?

Many thanks

Jonathan

Kostas Kyriakos · ‎09-26-2010

I think that the snmp-server host command tells the switch where to send SNMP notifications/traps to, rather than restricting SNMP polling to a specific host?

Yes that right. It sends traps to the host with ip SNMP_SERVER_IP.

@cweatherford:

Have you verified ip connectivity? What do the debugs indicate?

cweatherford · ‎09-26-2010

I can ping the switch from the server and I can ping the server from the switch. Here are the debug logs:

Sep 26 22:06:25.012: SNMP: Queuing packet to 10.110.1.191
Sep 26 22:06:25.012: SNMP: V1 Trap, ent ciscoConfigManMIB.2, addr 10.0.83.253, gentrap 6, spectrap 1
ccmHistoryEventEntry.3.106 = 1
ccmHistoryEventEntry.4.106 = 2
ccmHistoryEventEntry.5.106 = 3
Sep 26 22:06:25.264: SNMP: Packet sent via UDP to 10.110.1.191
Sep 26 22:06:27.990: %SYS-5-CONFIG_I: Configured from console by scvl on vty1 (10.110.1.191)
Sep 26 22:07:49.972: SNMP: Packet received via UDP from 10.110.1.191 on Vlan3
Sep 26 22:07:51.532: SNMP: Packet received via UDP from 10.110.1.191 on Vlan3
Sep 26 22:07:53.839: SNMP: Packet received via UDP from 10.110.1.191 on Vlan3

This is the config:

snmp-server community scvlc RO
snmp-server community scvl RW
snmp-server host 10.110.1.191 scvl

I tried removing the the last entry as well.

cweatherford · ‎09-27-2010

Any other thoughts?

cweatherford · ‎09-28-2010

I did find the following in the logs:

Sep 28 21:02:15.202: %SNMP-3-AUTHFAIL: Authentication failure for SNMP req from host 10.110.1.191
Sep 28 21:02:16.771: SNMP: Packet received via UDP from 10.110.1.191 on Vlan3
Sep 28 21:02:19.078: SNMP: Packet received via UDP from 10.110.1.191 on Vlan3
Sep 28 21:41:49.911: SNMP: Packet received via UDP from 10.110.1.191 on Vlan3
Sep 28 21:41:49.911: %SNMP-3-AUTHFAIL: Authentication failure for SNMP req from host 10.110.1.191
Sep 28 21:41:51.455: SNMP: Packet received via UDP from 10.110.1.191 on Vlan3
Sep 28 21:41:53.761: SNMP: Packet received via UDP from 10.110.1.191 on Vlan3

But my community names match.

cweatherford · ‎09-28-2010

Fix:

In NCM the global settings were not getting down to the node level. I had to select groups and then 'edit multiple nodes' and insert the correct community there...Here's my sign!