On Fri, Apr 26, 2013 at 10:38:52AM -0400, Chris Evich wrote: >On 04/25/2013 05:01 AM, Mike Qiu wrote: >> 2013/4/25 16:44, Wei Yang wrote: >>> All, >> +cc to [email protected] >>> I am confused by the behavior in migration test. Your clarification >>> will be >>> appreciated. >>> >>> In VM.migrate() for qemu_vm.py, it will try to save two guest to file and >>> compare the hash result of them. >>> >>> I met two issues: >>> 1. save_to_file() requires the guest to be in paused state. >>> While the current situation is one of the guest after migration is in >>> running state. >>> This is not a big deal, we can set the guest in paused state. >>> >>> 2. it will save the status to a file and compare them >>> In my mind, the state will be different. Since one guest is >>> continuing the >>> job and another one stops. >>> I am not sure in which case this comparison succeed? >>> >>> I am not familiar with those process. If I make a mistake, please let >>> me know. >>> >>> Thanks :) > >Wei, > >Mike moved this over to virt-test-devel list which is a better place. >Unfortunately lmr is out on PTO this week, and he would probably have a >better answer for you. However I can offer my experience in this area.
Chris, Thanks for your explanation. > >When I did some work on qemu_vm.vm.migrate, I also got stuck on many >things like what you mention. Eventually what I learned is, the >vm.migrate() method is only a small part of the picture. > >Also, qemu migration testing is a _very_ old code, there's likely a lot >of extra "stuff" that's been added on over the years for one reason or >another. What you found is probably there to support one aspect of >migration testing in a certain way. > >You can probably discover the reason by looking at the larger-picture: >actual migration test-modules, the autotest server-control files, >utils_test.MultihostMigration, and autotest.client.shared.syncdata. >They are all part of the big picture. Thanks, let me take a look into those files :-) > >If after you still can't find an answer after looking at the wider >code-paths, then come back to this list. Hopefully lmr will be back by >then also :) > >-- >Chris Evich, RHCA, RHCE, RHCDS, RHCSS >Quality Assurance Engineer >e-mail: cevich + `@' + redhat.com o: 1-888-RED-HAT1 x44214 -- Richard Yang Help you, Help me _______________________________________________ Virt-test-devel mailing list [email protected] https://www.redhat.com/mailman/listinfo/virt-test-devel
