On Wed, 20 Sep 2006 08:53:08 +0200, Rob van der Heij <[EMAIL PROTECTED]> wrote:
>On 9/20/06, Ranga Nathan <[EMAIL PROTECTED]> wrote: > >> Strangely some guests show individual address of hipersockets triad >> while configuring network via YaST in SLES9. For example, see below: > >> Again, it is only for some guests. > >Have you preserved the even/odd numbering when you map the real QDIO >device onto a virtual one? IIRC the driver in Linux arranges the >(virtual) devices in the order that fits the tripod arrangement, but >when you swap even and odd devices in the attach, the driver got them >wrong. >If you simply took the next 3 real devices and attach them as >EC00-EC02 for each server, that would suggest half of the guests get >them wrong. > >There's various ways you can ensure the even/odd mapping is retained >(including a virt=real mapping, skipping one address between the >tripods, alternating swap of the first two, etc). I'm coming late to this discussion due to being away at a DR exercise, bu t I don't match up even/odd address and have no trouble with any of the 13 Liunx guests using Hipersockets. Our Hipersocket addresses are attached to 13 different LINUX guests as follows: 713-715 attached as 703-705 716-718 attached as 703-705 719-71B attached as 700-702 71C-71E attached as 700-702 71F-721 attached as 700-702 722-724 attached as 703-705 725-727 attached as 703-705 728-72A attached as 703-705 72B-72D attached as 703-705 72E-730 attached as 700-702 731-733 attached as 703-705 734-736 attached as 700-702 737-739 attached as 703-705 All these LINUX guests are happly using their Hipersocket connections. There's lots of even/odd mixing going on in the above, so maybe I misunderstand your concern. We're running z/VM 5.2 and this has worked with the Linux guests at both SLES 8 & SLES 9. Brian Nielsen
