HP OfficeConnect 1920 ARP Problems

[lverhoeven]

I'm trying to setup my home lab using some Intel NUC's and an officeConnect 1920 switch. Because it's only a small network the 1920 should be enough to be used as a core switch.

I've connected four Intel NUC's directly to the switch in the default VLAN. My DHCP relay agent is in another VLAN which is also used for routing towards the internet. The NUC's are all running Centos 7 and are configured with a static ip. The problem is that I'm unable to connect to any other VLAN from the NUC's but they can all reach eachother without problems. When running tcpdump on one of the NUC's and starting a ping from the switch towards the NUC I can see that an ARP request is sent and replied to but the switch seems to ignore it. When looking at the ARP Managent page I can't see any of my NUC's.

When I configure a static ARP record I'm able to connect to the Internet. This is only a quick fix because it should work as expected and use dynamic ARP. I've had the same problem on my NAS, which is running FreeNAS, and solved it on this system by using DHCP. This is unfortunatly not an option on the NUC's.

Below is an output from my tcpdump command, I've confirmed that it reaches the switch by using port mirroring:

21:26:41.518185 ARP, Request who-has 192.168.11.92 tell 192.168.11.1, length 46
21:26:41.518228 ARP, Reply 192.168.11.92 is-at f4:4d:30:66:09:29, length 28
21:26:43.720976 ARP, Request who-has 192.168.11.92 tell 192.168.11.1, length 46
21:26:43.721019 ARP, Reply 192.168.11.92 is-at f4:4d:30:66:09:29, length 28
21:26:45.930901 ARP, Request who-has 192.168.11.92 tell 192.168.11.1, length 46
21:26:45.930942 ARP, Reply 192.168.11.92 is-at f4:4d:30:66:09:29, length 28
21:26:48.140938 ARP, Request who-has 192.168.11.92 tell 192.168.11.1, length 46
21:26:48.140979 ARP, Reply 192.168.11.92 is-at f4:4d:30:66:09:29, length 28
21:26:50.350895 ARP, Request who-has 192.168.11.92 tell 192.168.11.1, length 46
21:26:50.350935 ARP, Reply 192.168.11.92 is-at f4:4d:30:66:09:29, length 28

Is there anyone that can tell me how to fix this?

Thanks in advance.

With kind regards,

Luc

[Vince-Whirlwind]

I'm not even sure what devices are involved in that ARP exchange.

What IP address is on the NCU? And what default GW and subnet mask?

What IP addresses are on the switch? What subnet masks? What is the default GW/default route on the switch?

[lverhoeven]

Default VLAN:

• Switch: 192.168.11.1/24
• NAS: 192.168.11.10/24
• NUC1: 192.168.11.90/24
• NUC2: 192.168.11.91/24
• NUC3: 192.168.11.92/24
• NUC4: 192.168.11.93/24

Internet VLAN:

• Switch: 192.168.254.1/24
• Router: 192.168.254.254/24
• DHCP: 192.168.254.253/24

default route in the switch has a next hop of 192.168.254.254

[parnassus]

Can you also share sanitized HPE OfficeConnect 1920's running configuration?

[Vince-Whirlwind]

That seems very odd. There's something I'm not quite getting, though - why would a device that has obtained an IP address from DHCP behave any differently? Whether a device is DHCP-addressed or not makes no difference to how ARP works.

Why is DHCP not an option for the NUCs anyway? It's far better to centrally manage your IP addressing than have to do it manually in 5 different places.

[lverhoeven]

I'll try to upload the running config tonight.

The NUC's are running as hypervisors, the DHCP server is running on one of them. The virtual machines don't seem to have a problem with connecting to the switch at the moment. I've got no idea why dhcp seemed to fix the problem on the NAS, I just noticed this morning that after rebooting the switch that it's not able to connect to the switch again.

When I look at the MAC table all the mac addresses are present but they don't appear in the ARP table. I seems like I'm reaching a limit in the ARP table but there are only about 10 dynamic ARP records.

[lverhoeven]

The running config is available at http://pastebin.com/cy3k8EwE

The NUC's are connected on port 1 to 4, the nas is connected on port 8.

My router and dhcp are running on the NUC that is connected to port 1.

[Vince-Whirlwind]

Asymmetric routing caused by NUC1 being in both networks? It looks like NUC1 has IP addresses in both VLANs.

The ARP you see in your capture isn't the switch, it's NUC1. Check out NUC1's ARP table.
The thing is, a packet reaching 254.254 addressed to 11.90, the device is going to go, "ah ha, I have a leg in the 11 network, let me send an ARP request". The route you put on it is going to be ignored if there is Layer2 connectivity.

 Here's a test: change NUC2, 3, &4's default gateway to 11.90 instead of 11.1 and see what happens.

Your initial description of your design looked exactly how I like to configure networks, but your latest revelation about NUC1 is definitely not the way to do it. Also, you shouldn't use "hybrid" - no other vendor has such config, it's something peculiar to where the 3COM engineers are from and has no application in the normal world.

[16again]

Seems like the HP1920 switch doesn't like MAC address f4:4d:30:66:09:29, and thus doesn't add it to its ARP table.
Shut down and disconnect the NUC with that address, get another device where you manually can set the MAC address, set it to NUC MAC address, and see if it now also has problems reaching the switch.

MAC addresses f4:4d:30:66:09:29 is an Elitegroup device and seems valid though.

[lverhoeven · Accepted solution]

I've managed to solve it by moving away from the default vlan.

I've created a new vlan (11) and set this as the untagged vlan and PVID on all ports, I've had no problems since.

Thanks for all the help.