Hello,

We have this same issue at CSUEB. We have X amount of resources and X
amount of limited local storage space on our current Blade hardware.

For example, we have X core blades at X mhz per core with X gb of ram
and X GB HD of local storage that is not upgradable on our current
hardware. I have then calculated X number of VM's can run with X
amount of performance to failure ratio.

We are running 20 to 22 VM"s per blade that is about 1536mb of ram
each and we give each VM tmp directory about 1 to 3 gb of space per
VM. This makes each VM able to have 1 core at 2400mhz per VM, with our
current hardware. We have our Vmware tmp space set to /dev/shm and
have giving ourself 87 gb. We had it set to 40 GB but were also
running out of space. We have the main images and VMguest load on to
our netappfiler02:/vol/sata1/virtualmachines .

Here is the our solution for changing the local VMware TMP directory
on our blades

-Add to /etc/fstab:
none /dev/shm tmpfs size=87G 0 0 (You can set yours to whatever)
-Run 'mount -a'
-Add or edit the line in /etc/vmware/config:
tmpDirectory="/dev/shm"
-Restart VMware: 'service vmware restart'
-Restart the VMs on the host
-Verify that the memory-mapped files are being stored in /dev/shm:
'lsof | grep vmware-root'
-You should see the space used on /dev/shm increase.  Verify this by
running 'df /dev/shm'.

We are running a NFS share with our VCL, I also want to point that
out. We have tried to use the nFS share as the tmpDirectory for
VMware, this lead to failures and very slow speeds, the Vmware
tmpDirectory should be hosted local if you have the space.

I have noticed in testing as more software or data is loaded the VM's
increases in size on both the copy image (Vmguest-XX) this is storage
on our local NFS mount and the local blade VM directory. In campus
testing we have found out the hard way with this method due to
failures poor performance when reloading mutipal VM's on our small
amount of blades. We had originaly set our system to run 30 VM"s per
blades but ran out of resources and local storage space. When the
system reloads 30 to 50 + VM's it's going to  pushing X amount of
network space/local storage ect. If you run out of tmp space on your
local blade you get all failures on all reservations once your hit the
limit. Also when you reload 10 or 15 VM's at one time on a blade, It
can get very slow. We don't have 100's of blades like NC state, to
prevision and ease the load on our hardware, we had to cut the number
of VM's per blades to increase performance. To help with consistence
reloading, we have also changed some of the Virtual Environments with
4 or 5 softwares into one image; so multiple classes can use the same
image and not have to reload a image for each class. This has helped
tremendously here on our campus.

Here is an example of our local storage setup on one of our Blades
here. We are also using VCL 2.1 and Vmware 1.X with our older setup.

[root@vclesx04 ~]# df -h
Filesystem            Size  Used Avail Use% Mounted on
/dev/sda2             7.8G  3.6G  3.8G  49% /
/dev/sda6             3.8G  571M  3.1G  16% /var
/dev/sda5             3.8G  188M  3.5G   6% /home
/dev/sda1              97M   18M   75M  19% /boot
tmpfs                  87G   36G   52G  42% /dev/shm
netappfiler02:/vol/sata1/virtualmachines
                      5.3T  2.2T  3.1T  42% /virtualmachines

-Alex


On Tue, Apr 19, 2011 at 9:26 AM, IBRAHIM MOMOH <momohibra...@gmail.com> wrote:
> hi Andy,
>
> Thanks for your quick response.  Here are the requested details:
>
> vm guest files are taking up all the space. i think they should be deleted 
> after a session ends.
> Provisioning module: Vmware server
> VCL Version: VCL2.1
> VMware version: 1.0.10
>
>
> On Apr 19, 2011, at 10:05 AM, Andy Kurth wrote:
>
>> Hello,
>> More information is necessary.  Which files are taking up the space? Which 
>> provisioning module are you using?  Which version of VCL and VMware are you 
>> running?
>>
>> -Andy
>>
>> On 4/19/2011 8:15 AM, momoh Ibrahim wrote:
>>> Hi Everyone,
>>>
>>> My university have VCL (using vmware) running on 7 blades, it is
>>> difficult to make reservations because the blades are out of space. I
>>> have a feeling something is not being deleted after reservations are
>>> expired/completed. Any suggestions on what to do?
>>>
>>> Thank you.
>>>
>>>
>>> Ibrahim Momoh
>>>
>
>



-- 
Thanks,
Alex  Patterson
User Support Services
Operating System Analyst
California State University, East Bay

Reply via email to