Wired Intelligent Edge

Hello,

In my client we are experiencing a strange behaviour at zl8212.

Switch is working as a layer 3 routing switch.  

For some of the computers directly connected to a UTP port on zl8212 arp address resolution does not work as expected. Therefore I cannot ping directly from the switch cli. However computers in the same vlan can reach the Computer without any problem. 

Since routing is expected from the 8212 other subnets cannot reach the Computer neither.

When I look at the arp table of the 8212 I see incomplete arp entries. 

For example; "192.168.100.123   000000-000000"

Whenever I reach the Computer from an other computer at the same lan and ping to default gateway  (8212) all other nvlans start to reach the Computer.

I believe that there is an arp learning issue with the 8212 but I could not diagnose the problem. 

Does anybody have a similar exprience and any possible solution idea?

P.S. No! I did not reboot the switch yet. Since there is no redundancy in comtroller it requires planned down time. 

 The Computer at hand is running some king of stripped linux.

8212 Model number is J9091A - and firmware version is K15.04.07

Gökhan Alganatay

What you are describing would indicate that the devices are either

 - not receiving the ARP requests from the router

 - not responding to the ARP requests from the router

As you say everything starts working fine as soon as you "wake up" the computer by getting it to ping its default GW, that would indicate there is probably not a VLAN misconfiguration or hardware issue causing the problem and that the problem is with the host, probably something to do with the way its NIC is configured.

"computers in the same vlan can reach the Computer without any problem. "

This tells you you need to look at the NIC configuration: the host is happy about its Layer2 address, but not too clear on its Layer3 config.

Maybe the NIC is patched to a switchport that has multiple VLANs on it and the ARP response is being sent to the wrong network?

Maybe the NIC thinks it's using a different IP address?

Unfortunately there is no other active hardware between the Computer and the 8212. There is only one small piece of utp patch cord, which has already changed a couple of times to double check.

The port has only one vlan assigned and it is untagged. Everything is at their default values considering the mac learning.

It sounds like you have a duplicate MAC address on the network. I have seen this exact behavior and it turned out to be a duplicate MAC. When you ping from the device you are trying to reach, it resets the switch MAC and ARP tables but then after a while, the other device changes them back.

Look at the MAC tables when the problem is happening and search for the MAC address. I think you will find it attached to another port.

HP Switch# show mac-address 000352-09a236

Status and Counters - Address Table - 000352-09a236

Port VLAN
----- ----
1       1

To keep the metter simple I just focused on a single device. However there are a few devices dealing with this problem.

When I look at the arp table, all of these devices is listed with 000000-000000 as the mac address. 

On the other hand, starting from your suggestion the problem could be originated from mis-configured redundant paths. If the backbone sees the same mac adress comming from different ports It might get confused. But the device at hand is directly attached and this eliminated the possibility of two redundant links.

I'm at the starting point again :(

Gokhan

An all 0's mac is usually indicative of a unreachable next hop gateway, not sure that is the case here though.  Typically this means the next hop IP is not responding to an ARP request from the switch, as was mentioned in the other replies.  Can you attach the config?

I would try and verify if the ARP request/reply is taking place via packet capture on the host.  If the ARP request from the switch is being replied to then it could be a config or possibly a software issue.  K.15.04 is fairly old though I don't see anything similar in our bug database.

Hope that helps.

My client is not willing to share the configuration due to security concerns. However I'll investigate the issue according to your comments. Thanks in advance.