System SNMP error - SNMP::get(): No response from <remote host>

[Osiris]

Sorry, I wasn't reading closely enough. Does snmpwalk work to all the switches from the command line on the cacti server? If not make that the first thing to fix. Many HW vendors have ACL capsbilry on their switches, or it could be the mtu between you and the routers that Are not responding.

[CptnBlues63]

Sorry, I wasn't reading closely enough. Does snmpwalk work to all the switches from the command line on the cacti server? If not make that the first thing to fix. Many HW vendors have ACL capsbilry on their switches, or it could be the mtu between you and the routers that Are not responding.

Yes, it worked from the cacti server.

[TheWitness]

are you trying snmpwalk directly from cacti server?

If no, It could be routing issue (missing route or default gateway on switches).

If yes, try to delete these 3 switches from cacti and add it again. I met this issue sometimes.

Oh boy do I feel sheepish now..........LOL

I was doing that snmpwalk from my switch management server (also linux). I just tried it from my cacti server and it's not getting a response from those three switches. It is from the other switches.

I've just looked at the configuration and it's the same as the other switches. They have an IP, subnet mask and default gateway, all correct. We didn't bother putting routes on these because they're all on the same subnet which is isolated from the rest of our network.

All 3 switches respond to ping requests from the cacti server.

So what should I look at now?

So, they respond to ping, and can be responded to from your Management Server. So, I suspect the switches have an ACL that is blocking clients that can use SNMP by IP Address or by IP Address range. I'm not a network guy, so I can not tell you where to find it, but I know enough that this is more than likely the issue. If you can confirm that the switch is listening on TCP:161 too, that might help. If it's listening on TCP:161, try using the TCP protocol vs. the UDP protocol from Cacti. I suspect it's an ACL on the switch though.

[TheWitness]

Another thing I would do is set up a tcpdump from the Cacti server before running the snmpwalk. If you get an 'unknown port' response on UDP 161, that means that you are being blocked somewhere between you and the switch.

[TheWitness]

Another one. Setup a span port on the switch itself that echos the inbound management interface to your local sniffer and then see if the Cacti servers UDP traffic is reaching the switch, or setup the span port anywhere in the path, as close to the switch as possible. I use Wireshark now, but when I was in the business, I always used Network General. That was decades ago.