If you have some XenServer 5.6 host that randomly (and suddenly) becomes completely unresponsive with no apparent reason, then it might be affected by an odd bug that has been documented by Citrix.

Is my server affected?

Potentially, any server based on the Nehalem (Xeon 75xx, 55xx, 35xx, 34xx) and Westmere (Xeon 56xx, 36xx) architectures can be affected.

How to know if my XenServer 5.6 is affected?

1. Open a console to your XenServer host (SSH client, the console tab in XenCenter, or the local console).
2. Make sure that your server is running on any of the affected architectures. To find out, run:
   
   # cat /proc/cpuinfo | grep -m1 "model name" model name      : Intel(R) Xeon(R) CPU           E5606  @ 2.13GHz

   1 2	# cat /proc/cpuinfo | grep -m1 "model name" model name      : Intel(R) Xeon(R) CPU           E5606  @ 2.13GHz

3. If it is,  with administrative privileges run:
   
   # xenpm get-cpuidle-states | grep -m1 "total" total C-states       : 4

   1 2	# xenpm get-cpuidle-states | grep -m1 "total" total C-states       : 4

If the value of total C-States is bigger than 2, then this server is affected by the bug and if necessary, they must be disabled from the BIOS.

How to fix?
Continue Reading »