On Wednesday, 06/04/2008 at 12:58 EDT, Richard Clapper <[EMAIL PROTECTED]> wrote: > Here's an update on the question I initiated May 7th. In working with IBM, > wev'e gotten to the point of tracing messages from the external LAN coming into > the OSA, through the VLAN-aware VSWITCH, getting to a Linux guest, and the > reply going back to the VSWTICH and to the OSA. But a sniffer placed between > the OSA and its corresponding real switch shows no messages coming out of the > OSA. Everyone agrees the Linux routing and the VSWITCH definition look correct. > > Anybody ever experience anything like that? What would you do to analyze > and/or resolve that?
I'm assuming this is a Layer 3 (IP) virtual switch: 0. Double-check that the OSA you're using is the one you think you are using and that it is plugged into the correct switch port. 1. Use OSA/SF or OSA Facilities on the HMC to examine the card. Look for any anomaly. You're looking for a bogus registration of the Linux's default IP gateway address. That would cause the card to suck up the packet. 2. Issue NETSTAT ARP * against the controller virtual machine. You're looking to see if the MAC address of the gateway IP is in there. Is it correct? 3. Reset the OSA. (Vary off the chpid to all LPARs, then bring it back online.) 4. Open a problem with OSA support so that they can see whether they put it on the wire Alan Altmark z/VM Development IBM Endicott