2009-01-29

2009-01-27

Converting a Linux Virtual Machine With an LVM

I was entrusted with the task to try and migrate a VM from Xen 3.2 to ESX. Well you would think that is a trivial task - it should be. But let us not forget that we are talking here about version 3.2, which is two generations back from the current Citrix Xensource which is in use today. After successfully migrating it I would like to share with you the procedure. This was done on a Xen Guest VM runnning Red Hat Enterprise Linux ES release 3 (Taroon Update 7) Disk Layout was two disks, hda and hdb. The second disk was configured as part of a an Logical Volume as you can see below
[root@xxxx root]# fdisk -l Disk /dev/hda: 26.8 GB, 26843545600 bytes 255 heads, 63 sectors/track, 3263 cylinders Units = cylinders of 16065 * 512 = 8225280 bytes Device Boot Start End Blocks Id System /dev/hda1 * 1 61 489951 83 Linux /dev/hda2 62 322 2096482+ 82 Linux swap /dev/hda3 323 453 1052257+ 83 Linux /dev/hda4 454 3263 22571325 f Win95 Ext'd (LBA) /dev/hda5 454 1236 6289416 83 Linux /dev/hda6 1237 1497 2096451 83 Linux /dev/hda7 1498 1758 2096451 83 Linux /dev/hda8 1759 2019 2096451 83 Linux /dev/hda9 2020 2280 2096451 83 Linux /dev/hda10 2281 3263 7895916 83 Linux Disk /dev/hdb: 26.8 GB, 26843545600 bytes 255 heads, 63 sectors/track, 3263 cylinders Units = cylinders of 16065 * 512 = 8225280 bytes Disk /dev/hdb doesn't contain a valid partition table [root@XXXXXX root]# lvscan lvscan -- ACTIVE "/dev/OracleVG/lvol1" [10 GB] lvscan -- ACTIVE "/dev/OracleVG/lvol2" [5 GB] lvscan -- 2 logical volumes with 15 GB total in 1 volume group lvscan -- 2 active logical volumes
Fired Up VMware Converter version 4.0 Beta and started the import as if the machine was a powered on physical machine Source info Destination info And when we get to the options you as you can see below the Data to copy is missing a drive We continue with process of the import. The way the new converter works with a live Linux machine - it uses a helper VM. What it does is, it creates a VM on your ESX, powers it up and loads a Linux ISO (don't know which one) to act as a mediator for the cloning process. This VM has to be able to receive an IP from a DHCP server otherwise the process will not work. This is a picture of what is hapenning inside the console of the helper VM: Time for conversion - Load on Host - Completed The configuration of the new VM on ESX as you can see only had one drive. Powered on the VM and this is what I got. The OS would not come up because it was missing the second part of the volume. Entered the Root Password
vi /etc/fstab
Uncommented the mount points for the LVM and restarted.
/dev/OracleVG/lvol1 /oradata/[folder] ext3 defaults 1 2 /dev/OracleVG/lvol2 /oraback/[folder] ext3 defaults 1 2
Kudzu kicked in to recognize all the new hardware, LSI Scsi card, NIC and VGA. Machine came up GUI and all - like it should have. But I was missing the 2nd drive of course.
[root@XXXXX root]# fdisk -l Disk /dev/sda: 26.8 GB, 26863764480 bytes 255 heads, 63 sectors/track, 3266 cylinders Units = cylinders of 16065 * 512 = 8225280 bytes Device Boot Start End Blocks Id System /dev/sda1 * 1 61 489951 83 Linux /dev/sda2 62 322 2096482+ 82 Linux swap /dev/sda3 323 453 1052257+ 83 Linux /dev/sda4 454 3263 22571325 5 Extended /dev/sda5 454 1236 6289416 83 Linux /dev/sda6 1237 1497 2096451 83 Linux /dev/sda7 1498 1758 2096451 83 Linux /dev/sda8 1759 2019 2096451 83 Linux /dev/sda9 2020 2280 2096451 83 Linux /dev/sda10 2281 3263 7895916 83 Linux [root@XXXXX root]# lvscan lvscan -- no volume groups found
Added a 2nd Drive identical to the first and rebooted and got my second drive
[root@XXXXXX root]# fdisk -l /dev/sdb Disk /dev/sdb: 26.8 GB, 26865020928 bytes 255 heads, 63 sectors/track, 3266 cylinders Units = cylinders of 16065 * 512 = 8225280 bytes Disk /dev/sdb doesn't contain a valid partition table
Now comes the fun part: [source_ip] = old Linux Machine From the Destination (i.e. the new VMware VM we just created) run the following command:
ssh root@[source_ip] "dd if=/dev/hdb" | dd of=/dev/sdb
And then wait.... The only thing you will see is on the source that sshd is using resources, no other indications.
PID    USER PRI NI SIZE RSS   SHARE STAT %CPU %MEM TIME CPU COMMAND 16482 root  25  0   2152 2152 1628 R 13.1 0.2                0:03 0     sshd
Once that was done (took about 2.5 hours for a 20gb Disk) Got this in the end
52428800+0 records in 52428800+0 records out 52428800+0 records in 52428800+0 records out
Removed the comments that I made previously in /etc/fstab and another reboot And up she goes...
[root@XXXXX root]# fdisk -l Disk /dev/sda: 26.8 GB, 26863764480 bytes 255 heads, 63 sectors/track, 3266 cylinders Units = cylinders of 16065 * 512 = 8225280 bytes Device Boot Start End Blocks Id System /dev/sda1 * 1 61 489951 83 Linux /dev/sda2 62 322 2096482+ 82 Linux swap /dev/sda3 323 453 1052257+ 83 Linux /dev/sda4 454 3263 22571325 5 Extended /dev/sda5 454 1236 6289416 83 Linux /dev/sda6 1237 1497 2096451 83 Linux /dev/sda7 1498 1758 2096451 83 Linux /dev/sda8 1759 2019 2096451 83 Linux /dev/sda9 2020 2280 2096451 83 Linux /dev/sda10 2281 3263 7895916 83 Linux Disk /dev/sdb: 26.8 GB, 26865020928 bytes 255 heads, 63 sectors/track, 3266 cylinders Units = cylinders of 16065 * 512 = 8225280 bytes Disk /dev/sdb doesn't contain a valid partition table [root@XXXXX root]# lvscan lvscan -- ACTIVE "/dev/OracleVG/lvol1" [10 GB] lvscan -- ACTIVE "/dev/OracleVG/lvol2" [5 GB] lvscan -- 2 logical volumes with 15 GB total in 1 volume group lvscan -- 2 active logical volumes
and of course GUI and all

2009-01-23

Exporting Virtual Machines to another computer

A potential client (hopefully will be a major client in the future) came to me with a requirement to consider ESX3i as their next platform for a stand-alone lab (x30). The major setback at the moment was the ability to easily export VM's from one Lab to another. They needed to exprt the image off of host relatively often. Now why they were not using ESX in the first place and why they are continuing not to do so, and why they would not go for an enterprise solution, that is a whole different story, which I will not get into. And neither why ESX does not really compare to Xensource.

They are using Citrix Xensource 3.2 - which is a quick Right-click -> Export ... Choose your location -> Save.

Since each VM is something like 40gb in size, this can be a very time consuming process.

So they started out by trying with thick disks (40GB) and to transfer that kind of a disk over the network can take mmmmm.... quite a while..

Not acceptable!

They went for Thin-disks (yes - I know that it sucks performance-wise..). Well when we created the disks it showed the size of 40GB on the VM's settings, when ls -la on the Console (yeah yeah not supported - I know..) but when looking at it through the Datastore Browser - lo and behold - 8GB in size, exactly what we were expecting.

So we tried to move the VM off of the Host, with the Virtual Infrastructure Client GUI, with WinSCP, with ssh, with FastSCP, we all go the same results (as was expected), as soon as the files was exported off, it grew to 40GB in size, and therefore took too long to export the VM.

Again Not acceptable!

So what was left, and I have Edward again here to thank for reminding me, the export feature.

Aha!!

In the VIC, when the machine is powered off, when you select your VM and then from the top Menu, File -> Virtual Appliance -> Export, you have the option to Export the VM out to any location accessible from your computer. Me be the skeptic I am, I thought that it would not make much of a difference in size and time, but I was wrong.

Test 1: Export VM with 40GB Thin disk to local C: drive
Time to completion: 12 minutes
Size of OVF: 1.2 GB

Test 1: Export VM with 40GB Default Thick disk to local C: drive
Time to completion: 15 minutes
Size of OVF: 1.7 GB

So not only do they have their export feature, not only are the VM image sizes manageable, not only is the timeframe more than acceptable, the disks will stay Thick by default which I am sure will improve the performance, and also make Edward happy.

So, perhaps I have convinced the client that ESX3i is more than a suitable and worthy candidate that will match and even improve on that feature from Xensource any day (not to mention almost any other feature as well). And if all goes as I hope, we will have another happy client that will benefit from using ESX, and in the future they will go with the product they are supposed to be using the first place - VMware Lab Manager.

2009-01-18

My new ESX3i Whitebox

So I finally got around to installing a box for myself to play around with at work.

Of course I do not want a noisy server running under my desk (I am sure that my co-workers in the room would not be too happy with it either..) so I had to go for something quieter..


Lo and Behold my new Whitebox!!

HP DC 7800 Desktop with 8gb RAM 250gb HD Sata drive, running ESX3i and as quiet as a kitten.

Price-wise:

HP Dc7800 DC E6550 - $630
4x2gb Kingston KVR800D2N5 - $176
Intel Pro 1000/MT - $30 (that is because the on-board Intel NIC refused to work!)
250GB HD - $60

Total for the Whitebox $900 (But in all honesty, since I work as part of the IT group, the only thing that was actually purchased was the Desktop, all other hardware was provided through our department)

I have already put up 3 machines on it, two that I work with daily (1 Ubuntu,  1 Windows 2003 Server) and the 3rd one is a test for Windows server 2008 R2 Beta which I am trying out.




Thanks to Eric Sloof for the inspiration for the post. (did not come out as cheap as his though..)

2009-01-17

Veeam FastSCP 3.0 for ESX/ESXi Public Beta

It is finally out.

The biggest addition to this Beta is the fact that it works with ESX3i.

Get it here

2009-01-14

RoundTable was a no-go tonight - Hold on it happened in the end

Due to unforseen technical difficulties (Talkshoe messed up I guess) the roundtable was lots of fun but not audio

Will see you all in a week!!

Chat was fun though...


----


Whoa!! - Hold on - it happened in the end started only 35 minutes late.

So we talked a bit about exchange on VMware - but this was just a small teaser for the proper thing next week.

Thanks John for working to get it going..

2009-01-12

Why are there two VMware VMTN communities with the same name

I came across something really weird today.

Vmware have lots of community forums, and while trying to make some order in the alerts I receive from the forums, I noticed that there are two communities that deal with the same thing, but are in different places, with different content.



and this one



Now I asked the group of Virtual folks on Twitter and the answers I have received so far is:

  • One is vCenter and the other is vCenter Server

  • Different Levels in the VMware Communities

But seriously folks.. Why have two different forums with almost the same content, we actually talked about this topic on the Communities Roundtable Podcast #27, so until then I am going to have to continue receiving alerts from both locations till VMware gets their act sorted out.

Edit:

the answer I got from John Troyer: 
@maishsk First URL is vCenter (Server), 2nd has no threads, is *parent* category for management tools (look at upper right for children)


Makes sense, but it is still confusing ??????????

2009-01-06

So how are you all doing with SCVMM managing vCenter ???

That was actually meant to be highly sarcastic, it seems that the folks in Microsoft thought this would be a good idea (acutally it would be very good - if it would work properly..)

So you can control vCenter with SCVMM but it seems that when you want to remove it, typically like microsoft does, it leaves a lot of "leftovers" behind. So I personally would like to have my environment back the way it was before Microsoft started to play with and Eric Grey posted on his blog what is left over when you remove SCVMM. He even went as far as to provide a Powershell script that will remove all the muck for you.

Thanks Eric!!!