TCPIP works really nicely using a VSWITCH... at my current location, we've reduced TCPIP's role to the VM stack only -- and share the VSWITCH with Linux guests. We can bounce TCPIP without affecting anyone but the VM sysprogs telnetting in..
I'm a fan of letting the vswitch controllers manage the real OSA's - and leaving VM TCPIP to manage the VM stack.. with virtual devices attached to a vswitch. The VSWITCH implementation seems really solid in my experience. Scott On Mon, Aug 24, 2009 at 5:12 PM, Alan Altmark <alan_altm...@us.ibm.com>wrote: > On Monday, 08/24/2009 at 02:19 EDT, "Dean, David (I/S)" > <david_d...@bcbst.com> wrote: > > We had a situation where one of our OSA links died, but failover did not > > > occur. It appeared to us that the zVM was not aware that the card was > actually > > ?down?. This seemed similar to what you were referencing. I could be > wrong?.. > > There can be failures where OSA does not notify the host of a failure. The > VSWITCH was designed to detect an OSA "stall". TCP/IP's OSD device driver > was not. > > I am rapidly moving to the opinion that VM TCP/IP should use a VSWITCH and > let the VSWITCH handle failover at the hardware level rather than needing > multiple IP addresses, VIPAs, and dynamic routing. > > When VSWITCHes first became available, I was reluctant to put all my eggs > in one basket. Now, however, "no problem." > > Alan Altmark > z/VM Development > IBM Endicott >