[Xen-devel] More force pushes (xen-4.5-testing, xen-unstable, qemu-mainline)

[Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx>]

Here are some more which I propose to force push.  In each case I show
all the blocking tests mentioned in the osstest test report.


osstest service user writes ("[xen-4.5-testing test] 50373: regressions - 
FAIL"):
> flight 50373 xen-4.5-testing real [real]
...
>  test-amd64-i386-freebsd10-amd64 14 guest-localmigrate/x10 fail in 50317 
> REGR. vs. 50268
...
> version targeted for testing:
>  xen                  0b754fb3ed6b7b6d0f2e1f7c1877d3c0a7da2168


osstest service user writes ("[xen-unstable test] 50390: regressions - FAIL"):
> flight 50390 xen-unstable real [real]
>  test-amd64-i386-freebsd10-i386 13 guest-localmigrate      fail REGR. vs. 
> 36514
>  test-amd64-i386-freebsd10-amd64 15 guest-localmigrate.2   fail REGR. vs. 
> 36514
...
> version targeted for testing:
>  xen                  123c7793797502b222300eb710cd3873dcca41ee


osstest service user writes ("[qemu-mainline test] 50391: regressions - FAIL"):
> flight 50391 qemu-mainline real [real]
...
>  test-amd64-i386-freebsd10-i386 13 guest-localmigrate      fail REGR. vs. 
> 36709
>  test-amd64-i386-freebsd10-amd64 15 guest-localmigrate.2   fail REGR. vs. 
> 36709
> version targeted for testing:
>  qemuu                58c24a4775ef7ccf16cfcd695753dfdf149fe29d


These are all FreeBSD migration tests.  These seem to be failing much
more often in the new colo.  As I say I don't know why this is.  We
hope to fix it with the FreeBSD 10.1 update which Roger has submitted
and which is currently in the osstest self-push-gate.

In the general case they fail after a number of successful runs.  In
each case the symptom is that the guest simply doesn't come back onto
the network after the one of the migrations.  I double checked
  
http://logs.test-lab.xenproject.org/osstest/logs/50391/test-amd64-i386-freebsd10-amd64/info.html
and as with the earlier tests there doesn't seem to be any hints in
any of the logs.

These are all localhost migrations.  This rules out a problem with the
colo infrastructure dropping or not honouring gratuitous ARP or MAC
address migration, since the guest's MAC and IP address remain (from
the infrastructure's point of view) on the same switch port.

Ian.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel