On 2018-10-18 16:53, Wei Huang wrote: > > > On 10/18/2018 07:43 AM, Emilio G. Cota wrote: >> On Thu, Oct 18, 2018 at 14:38:01 +0200, Thomas Huth wrote: >>> On 2018-10-17 21:28, Emilio G. Cota wrote: >>>> Can anyone reproduce this? Otherwise, let me know what other info >>>> I could provide. >>> >>> I've finally been able to reproduce it - seems like it only happens here >>> when the host is under heavy load. > > Given that it only happens under heavy load, could the failure be caused > by migration thresholds defined in migrate_postcopy_prepare()? You can > play with "max-bandwidth", "downtime-limit" and others to see if it helps.
David suggested on IRC yesterday that this problem might be the same as with the ppc64 migration test: It does not work with TCG, only with KVM. I've now done some tests, and indeed, I can not reproduce the problem when running with KVM on a s390x host. But as soon as I disable the kvm kernel module and run the test with TCG there, the problem occurs again. So I think I'll send a patch to disable this test with TCG for now. Once the problem with TCG has been solved, we can enable it again. Thomas