Bug 717882

Summary: BSOD occurs when start 4 128GB windows guest on 512 GB host
Product: Red Hat Enterprise Linux 5 Reporter: Mike Cao <bcao>
Component: kvmAssignee: Virtualization Maintenance <virt-maint>
Status: CLOSED WONTFIX QA Contact: Virtualization Bugs <virt-bugs>
Severity: medium Docs Contact:
Priority: high    
Version: 5.7CC: bcao, michen, mkenneth, pbonzini, rhod, virt-maint
Target Milestone: rc   
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2011-06-30 16:54:03 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Attachments:
Description Flags
screen dump
none
qemu-kvm command line
none
Bugcheck Analysis
none
host infomation none

Description Mike Cao 2011-06-30 10:04:45 UTC 
Description of problem:
start 4 guests to test 

Version-Release number of selected component (if applicable):
# uname -r ;rpm -q kvm
2.6.18-269.el5
kvm-83-238.el5


How reproducible:
sometimes 

Steps to Reproduce:
1.install windows 2008 R2 guest with -smp 4 -m 4G, on a small memory host
2.copy the origanal host into 4 copies. named win08.5 ,win08.6, win08.7,win08.8
3.start the 4 guests on a 512 GB host ,each guest starts with -m 128G -cpu 16.(commandLine will be attached )
4.running #stress in 4 guests
  
Actual results:
BSOD occurs in one guest ,other guests work well


Expected results:
no BSOD occurs.

Additional info:
Comment 1 Mike Cao 2011-06-30 10:06:25 UTC 
Created attachment 510623 [details]
screen dump
Comment 2 Mike Cao 2011-06-30 10:08:08 UTC 
Created attachment 510624 [details]
qemu-kvm command line
Comment 3 Mike Cao 2011-06-30 10:11:17 UTC 
Created attachment 510625 [details]
Bugcheck Analysis
Comment 4 Mike Cao 2011-06-30 10:14:26 UTC 
Created attachment 510627 [details]
host infomation
Comment 5 Paolo Bonzini 2011-06-30 10:28:08 UTC 
This is not really a bug; bugcheck 0x101 simply means that the system is overcommitted.  There is a Viridian extension that makes Windows suppress the watchdog, we do not have it implemented.
Comment 6 Ronen Hod 2011-06-30 16:54:03 UTC 
I take Paolo's word above, and close the bug. It is probably possible to do better, but the problem will always reveal itself (probably with more guests running in parallel). How is RHEL6 running this test?
Comment 7 Mike Cao 2011-07-01 02:15:15 UTC 
Hi, Paolo ,Ronen

Could you gave we a criteria that How much memory over-commit does we support in RHEL5 host ?
From this test ,I did 20G host memory overcommit ,and the host is 504G memory.

Best Regards,
Mike
Comment 8 Paolo Bonzini 2011-07-01 08:47:12 UTC 
It's not possible to give a number valid in all circumstances.  For Windows, the guest watchdog means that as soon as you overcommit all bets are off.  (Without overcommit, bugcheck 0x101 can still happen but if it is reproducible it should probably be considered a bug; see for example Xen bug 651912).