VC: Can't ping some hosts on the same subnet

[james_sr]

Hello,

We're testing our new architecture based on an enclosure  c7000 with a virtual connect module installed.  During our differents tests, we encountered a strange issue. From our machine,  we were not able to ping some hosts which are on the same subnet as our.

On a virtual connect Manager, we've defined two differents networks and it happened with both. Some hosts are unreachable.  And we have not ACL that could filter the different destination IPs for which it's not working. 

 I could see the frame arriving on the target and the  reply to the arp broadcast being sent.  But when checking on the source host, no trace of the reply.  It sounds it disappeared at VC level.

This is happening only in this case. Otherwise, if we tried to ping a machine located in another subnet : no problem. It's working well and with all hosts.

Did this happen to someone? Or do you have any ideas on this?

Thanks 

James.

[Casper42]

Do you happen to have the "Private [X]" option turned on for those Networks in VC?

[james_sr]

I've just checked to be sure but nope, it's not enabled. What is very strange is that it only happens with particular hosts for which we haven't a particul config. and OS is different.

Thanks a lot,

[JimTonic]

What happens if You exchange the ip adress of 1 working (answering the ping) and 1 not working (not answering the ping) host/system?

If behaviour changes (working to not working and vice versa) it is rather the "different OS" than the subsystem/infrastructure/switch/VC.

Are the non answering ips pingable from non VC related systems/IPs?

[james_sr · Accepted solution]

Hi, 

Thanks for you reply. Yes the IPs were pingable from non VC related systems/IPs. I was able to ping them from my machine. 

During our tests today, we figured out what was happening. In fact, it was linked to the configuration between the uplinks from the 2 VC modules and our Cisco switch. It was misconfigured.  I had a Port-Channel set up and I was in active/active mode across 2 VC scenario.  That's why I have this strange behaviour. You could find the details in another post I opened 

http://h30499.www3.hp.com/t5/HP-BladeSystem-Virtual-Connect/VC-mode-Active-Active-Host-unreachable-when-using-virtual/m-p/6207523#M3048 

But in short, because of this misconfiguration, I had the traffic not going through the uplink I had with the second virtual connect.  Moreover, I understood another important by reading the technical whitepaper "HP Virtual Connect the Cisco Network Administrator": 

" Broadcast, multicast, and unknown unicast (destination lookup failure) frames are handled

slightly differently than unicast frames. They are load balanced so VC does not always send
these frames down the first port in the channel. The algorithm used to determine the port in the
channel is based on the source and destination MAC addresses, the source port channel number
(if source is a port channel) and/or module ID plus port number (if source is a regular port and
not a port channel). " 

As I have the VC2 uplink not working, this above seems to explain why I wasn't able to ping some random machines of the same subnet. 

Thus, by changing the port-channel config. to  a trunk  as Casper suggested in the other post,  it fixed this one too. 

 As I thought it was a totally a different issue, I would never imagine that it could be related to the other post. That's why I opened a new one. 

Thanks again.