[Xen-users] Windows 2008R2 HVMs and clock drift

[Tod Liverseed <tglcode@xxxxxxxxx>]

I am running Xen 4.0.1 on debian squeeze. I have 2 Win2008R2 HVMs running on it.  Dom0 is using NTPDate and has a solid time. Zero problem with Dom0. Now,  the strange part.  One vm's clock drifts,  while the other vm's clock does not drift.  Of course I do use NTP as both are part of a windows domain,  so ultimately W32TM does correct the issue (or keeps it within the domain tolerance). The .CFGs are identical except for the ram/disks and... VCPUs

OK,  HVM configs. These settings are common in both the VM that drifts and the one that does not. 

acpi = 1
apic = 1
device_model = '/usr/lib/xen-4.0/bin/qemu-dm'
boot="c"
localtime=1
sdl=0
serial='pty'
timer_mode=2
viridian=1

Now for the Differences....

********** Does not Drift **********
builder='hvm'
memory = 16384
shadow_memory = 96
vcpus=6
cpus="8-15"
**************************************

********** Drifts **********
builder='hvm'
memory = 8096
shadow_memory = 30
cpus="4-5"
vcpus=2
****************************
Below is a /stripchart from W32TM in the one that drifts. 

16:39:02 d:+00.0625636s o:+03.8884581s  [                           |          *                ]
16:39:33 d:+00.0625636s o:+03.9052416s  [                           |          *                ]
16:40:03 d:+00.0624592s o:+03.9651983s  [                           |          *                ]
16:40:33 d:+00.0624592s o:+04.1260910s  [                           |          *                ]
16:41:03 d:+00.0624592s o:+04.3182343s  [                           |           *               ]
16:41:33 d:+00.0780740s o:+04.7369497s  [                           |            *              ]
16:42:03 d:+00.0624592s o:+04.9525362s  [                           |             *             ]
16:42:33 d:+00.0624592s o:+05.3478084s  [                           |              *            ]
16:43:04 d:+00.0624592s o:+05.7430806s  [                           |               *           ]
16:43:34 d:+00.0624592s o:+06.1071022s  [                           |                *          ]
16:44:04 d:+00.0624592s o:+06.7836298s  [                           |                  *        ]
16:44:34 d:+00.0624592s o:+07.0382848s  [                           |                  *        ]
16:45:04 d:+00.0624216s o:+07.2666513s  [                           |                   *       ]
16:45:34 d:+00.0624216s o:+07.6019390s  [                           |                    *      ]
16:46:04 d:+00.0624216s o:+08.1559809s  [                           |                     *     ]
16:46:34 d:+00.0624216s o:+08.5850204s  [                           |                       *   ]
16:47:05 d:+00.0637092s o:+08.6929941s  [                           |                       *   ]


I have two identical servers,  running these same HVMs,  and both do the same exact thing. Both servers are Supermicro's with dual Xeon L5630 CPUs and 64gig of ram. I can't help but think that the VCPUs has something to do with this... I really can't try upping the VCPUs as these are running in production/dr environments. 

Any ideas?

Thanks :-)

Tod





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