[Xen-users] Crash *always* after approximately 1024 migrations

[Fraser Campbell <fraser@xxxxxxxxxxxxxxxxxxxxx>]

Hi,

We are running Xen 3 guests on NFS.  One of our stress tests involves 
repeated live migration of a domU between 2 Xen hosts.  Here are the 
basic steps:
* start domU on hostA
* start script on hostB that will live migrate to hostA once it sees
  running domU
* start script on hostA that will live migrate to hostB once it sees
  running domU (this starts the sequence)

The test always fails at close to 1024 migrations away (i.e. guest 
itself would have been migrated 2048 times).  I say close because it 
never seems to quite get to 1024 and earlier runs crashed at 1018, 1020 
and 1023 ... tonight's run crashed at 1021.
The domU can be under heavy load or no load, it makes no difference.

One of the hosts will freeze solid, the guest will be migrated to other 
host but still in paused state.
Name                              ID Mem(MiB) VCPUs State  Time(s)
Domain-0                           0     1000     1 r----- 11728.2
toxenc03                         1022      512     1 --p---     0.0

I am hoping this is a known problem.  If it is not I will gather more 
information before and after each migration, document exact sequence of 
events and repeat twice to ensure some consistency.  If there are any 
special points of interest that I should gather please advise.
We are using BETA of both SLES10 and FC5.  Kernels have been both vendor 
supplied and custom compiled, behaviour the same across the board. 
Servers pairs were either 2x HP-DL380G4 or 1x IBM-x366 and 1x HP-DL380G4.
Thanks

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users