Bug 740221

Do we know how long a physical Windows 2003 server in that configuration would take to boot?

This BZ compares win2k8 and win2k3 we need to compare physical to virtual.

Does your hardware support nested paging?
Can you try without spice (w/ vnc)? We had issues in the past with mapping of the memory slots of spice. Lastly, win2k8R2 is more important than win2k3 which is very old

(In reply to comment #2)
> Do we know how long a physical Windows 2003 server in that configuration would
> take to boot?
> 
> This BZ compares win2k8 and win2k3 we need to compare physical to virtual.

I am finding machine with 512G memory.  i will update testing result to bz as soon as possible.

(In reply to comment #3)
> Does your hardware support nested paging?
> Can you try without spice (w/ vnc)? We had issues in the past with mapping of
> the memory slots of spice. Lastly, win2k8R2 is more important than win2k3 which
> is very old

Hi Dor
   physical host support nested paging, without spice have the same issue too. 
 In addition,win2008-64 have the same issue too. it take 1.5 hours to boot completely.

Marcelo, it looks it worth to test for 6.3 since its uses nested paging.

The host has 504GB memory, starting a 512G guest exceeds its RAM size. Since Win2003 clears memory on startup, a 512G guest on this host goes deep into swap.

Booting guest with 480GB takes 5 minutes.

Closing as NOTABUG.

(In reply to comment #11)
> The host has 504GB memory, starting a 512G guest exceeds its RAM size. Since
> Win2003 clears memory on startup, a 512G guest on this host goes deep into
> swap.
> 
> Booting guest with 480GB takes 5 minutes.
> 
> Closing as NOTABUG.

If I can reproduce this bug in 1T host then I need to reopen it?

(In reply to comment #12)
> (In reply to comment #11)
> > The host has 504GB memory, starting a 512G guest exceeds its RAM size. Since
> > Win2003 clears memory on startup, a 512G guest on this host goes deep into
> > swap.
> > 
> > Booting guest with 480GB takes 5 minutes.
> > 
> > Closing as NOTABUG.
> 
> If I can reproduce this bug in 1T host then I need to reopen it?

Yes, you can then reopen, but please make sure the host is not swapping (see vmstat -a 1 output or cat /proc/swaps).

Since still can reproduce it in 1T memory big machine. So re-open this bug. 

This are detail testing steps:

1. reboot host - (in order to let all applications to release memory)

2. check free memory and swap in host

# free -g
             total       used       free     shared    buffers     cached
Mem:           991          5        986          0          0          0
-/+ buffers/cache:          5        986
Swap:            3          0          3

# cat /proc/swaps(no use swap) 
Filename			Type		Size	 Used	Priority
/dev/dm-1                       partition	4194296	  0	-1


3. boot win2003-64 guest with 512G memory and 64 cores vcpus

result:
Guest will take >20 mins to boot.

host info:
cpu:
processor	: 159
vendor_id	: GenuineIntel
cpu family	: 6
model		: 47
model name	: Intel(R) Xeon(R) CPU E7- 2860  @ 2.27GHz

memory:
# cat /proc/meminfo
MemTotal:       1040136796 kB(1T)

If host is rhel6.4, please use -M rhel6.4.0

FuXiangChun,

Profiled Windows 2003 guest boot on the machine with perf, thanks for setting it up.

Windows is clearing all of the memory during bootup (512G). There is little KVM interference, so it should take approximately the same amount of time to boot Windows 2003 on real HW.

Unless it can be shown that on real hardware Windows takes significantly less time to boot with such large amount of memory, this ticket should be closed as NOTABUG.

(In reply to comment #20)
> FuXiangChun,
> 
> Profiled Windows 2003 guest boot on the machine with perf, thanks for
> setting it up.
> 
> Windows is clearing all of the memory during bootup (512G). There is little
> KVM interference, so it should take approximately the same amount of time to
> boot Windows 2003 on real HW.
> 
> Unless it can be shown that on real hardware Windows takes significantly
> less time to boot with such large amount of memory, this ticket should be
> closed as NOTABUG.

I will reserve a 512G memory machine(since hasn't 1T memory machine in local) to install windows 2003, and update testing result to bz.

Tested this issue on 512G bare machine. windows2003 os only took 2 minutes to boot. and os work well.  

host:512G memory
cpu:48 cores ,PowerEdge R815

Created attachment 713555 [details]
memory usage for win2003

Please retest with RHEL7 packages (or upstream). If the problem persists, we'll move this bug to RHEL7. If it's solved, we'll close it as WONTFIX in RHEL6 due to capacity.

I think RHEL7.0 still has this issue.  It took 8 minutes to boot guest. If boot guest in 1T host. It will take more time.

Host: AMD 6172(Memory 512G, cpu: 48 cores)

# uname -r
3.9.0-0.rc6.51.el7.x86_64
qemu-kvm-1.4.0-2.el7.x86_64

Boot win2003 64bit guest with (free -g) memory
/usr/libexec/qemu-kvm -M pc-i440fx-1.4 -enable-kvm -m 504G -smp 4,sockets=2,cores=2,threads=1 -name windows -uuid 745fe449-aac8-29f1-0c2d-5042a707263b -nodefconfig -nodefaults -drive file=/home/win2003-64-virtio.qcow2,if=none,id=drive-virtio-disk0,format=qcow2 -device virtio-blk-pci,drive=drive-virtio-disk0,id=virtio-blk-pci0 -usb -device usb-tablet,id=input0 -vnc :0 -vga qxl -monitor stdio

(In reply to FuXiangChun from comment #32)
> I think RHEL7.0 still has this issue.  It took 8 minutes to boot guest. If
> boot guest in 1T host. It will take more time.
> 
> Host: AMD 6172(Memory 512G, cpu: 48 cores)
> 
> # uname -r
> 3.9.0-0.rc6.51.el7.x86_64
> qemu-kvm-1.4.0-2.el7.x86_64
> 
> Boot win2003 64bit guest with (free -g) memory
> /usr/libexec/qemu-kvm -M pc-i440fx-1.4 -enable-kvm -m 504G -smp
> 4,sockets=2,cores=2,threads=1 -name windows -uuid
> 745fe449-aac8-29f1-0c2d-5042a707263b -nodefconfig -nodefaults -drive
> file=/home/win2003-64-virtio.qcow2,if=none,id=drive-virtio-disk0,
> format=qcow2 -device
> virtio-blk-pci,drive=drive-virtio-disk0,id=virtio-blk-pci0 -usb -device
> usb-tablet,id=input0 -vnc :0 -vga qxl -monitor stdio

Is it possible to perform the test with -cpu host?

I　don't reserve AMD 6172 host. Maybe need to update result later.  I have contacted the current user of host. I will start to test this issue as soon as owner of host released machine.

(In reply to Ademar Reis from comment #31)
> Please retest with RHEL7 packages (or upstream). If the problem persists,
> we'll move this bug to RHEL7. If it's solved, we'll close it as WONTFIX in
> RHEL6 due to capacity.

Adermar,

Reproduced this issue with 1T memory & 48 cores on AMD machine (not found 1T memory amd machine before). According to comment31 & comment32 & comment37. For RHEL7.0 host, Is this issue handled the same as RHEL6 host?  Do QE need to clone this bug to RHEL7.0? 

Another, I tested several key points and checked host's memory usage.
1. Tested -M host & AMD machine 

2. Boot RHEL7.0 guest with -m 512G

3. free -g on host
# free -g
             total       used       free     shared    buffers     cached
Mem:           994        463        531          0          0         14
-/+ buffers/cache:        449        545
Swap:            3          0          3

4.# cat /proc/swaps 
Filename				Type		Size	Used	Priority
/dev/dm-0                               partition	4194300	0	-1

5.host info
processor	: 47
vendor_id	: AuthenticAMD
cpu family	: 21
model		: 2
model name	: AMD Opteron(tm) Processor 6344

6.hostname
hp-dl585g7-01.rhts.eng.nay.redhat.com

7. kernel & qemu version
3.10.0-121.el7.x86_64 & qemu-kvm-1.5.3-60.el7.x86_64