> On Friday, 01/25/2008 at 05:40 EST, [EMAIL PROTECTED] wrote: > > the switch IMLed and connectivity came back for AIX and PC servers - not > for > > the mainframe. > > first question: why doesn't VM TCPIP recover like AIX? > > I don't know what "like AIX" means, so I can't answer your question.
If a pSeries or xSeries adapter loses physical link for some reason, the interface appears logically down to the OS. When the link comes back, the interface sees the transition and removes the logically down state without human intervention. > > i have a VSWITCH set up with two controllers for failover and two OSAs > for > > physical failover. > > i see in the OPERATOR log where the VSWITCH tried to failover to the > other > > VSWITCH controller and then tried to failover to the other OSA. neither > worked > > because both OSAs lost network connectivity when the switch went down. > > > > what should be the recovery approach in this situation? > > Assuming you're on z/VM 5.3, do nothing. The VSWITCH will recover when it > is fixed. > > If you actually have an OSA problem that requires the offline/online reset > dance, as soon as the OSA goes away the VSWITCH will switch to a backup. > Of course, if you keep pulling all the OSAs out from under the VSWITCH, > there's not much it can do. When you get an OSA back online, issue a SET > VSWITCH CONNECT to re-establish the connection. Best is to take one OSA > out, bring it back, then another out and bring it back. Then you still > need do nothing. > > When you take the chpid offline, it's like unplugging the adapter from the > machine as far as the operating system is concerned. Does AIX actually > tolerate such a thing? If it's a hotplug adapter, yes, and certainly in an LPAR it works (that's what dynamic configuration and I/O partitions are for: isolating the physical adapters from the LPARs). If a device with the same characteristics comes back at the same IOvec (their analogue to cuuu), then AIX happily says "let's go" and does.
