Hi, My apologies to those of you that called in earlier unfortunately no one from OSDL was available to initiate the call so we had cancel it. Since in 2 weeks I'll be at the Japan meeting our next call will be Tuesday September 21, at 11:00am Pacific, 2:00pm East Coast. One of the things that I'd like to discuss at the next meeting is regression testing. Joel Schopp provided us with the following minimal configuration, please share your comments and suggestions on it: Joel Schopp wrote: "We run a number of cpu hotplug regression tests here at IBM. We'd be happy to donate them. We are of course targeted to ppc64, but many of the tests should run fine on other platforms. We basically have a stressor that flips cpus at random and a series of FVT (on the order of 15). Either can be run in combination with other tests such as LTP, ping flooding, parallel make, etc. The full FVT will require a 4 processor Power4 or Power5 pSeries machine in lpar mode and an accompanying hardware management console (HMC). This is a minimum requirement, more processors are fine. Also, for Power5 if tested with shared processors it should be noted only 4 virtual processors are needed, which can be done on a single physical processor. The stressor can be run on any platform that presents multiple (2 or more) cpus and supports cpu hotplug. i386 presents a dummy remove, but it good for exposing architecture independent bugs nonetheless. I am told that there are ia64 platforms that support cpu hotplug. Also, the 390 supports it as well. The stressor can of course also be run on pSeries lpar, though not in combination with the FVT." I've also started to write documentation for hotplug CPU that I'll be shortly submitting for review to this list before sending it to the sourceforge list. All input will be welcome. Thanks for your participation. Martine
