Thursday 30 August 2012

VMware vSphere Replication

Just 1 week ago I have finished building Disaster Recovery solution using vSphere Site Recovery Manager product. At this time vSphere Replication was solely part of SRM, but not of the vSphere. That's why I was a bit surprised when I saw "Introduction to vSphere Replication" as a part of "What's New in vSphere 5.1" documentation. At first sight it looked as a competitive contradiction to a current Disaster Recovery solution - SRM. But when I started to read the documentation I suddenly remembered that when I was playing around with Windows 2012 Release Preview I noticed that Microsoft included new free feature called Hyper-V Replica, which was supposed to provide Disaster Recovery functionality. Considering that it was absolutely free of charge benefit it had at least one strong advantage over VMware SRM.

So now it has gotten clear that this is just an answer to Microsoft challenge on Virtualization Market. When I went further into What's New documentation I found some other obvious evidence that VMware is striking back on Microsoft Server 2012.

Now, after I have done some reading I can see that vSphere Replication can replace SRM in small companies only. Here are the reasons:
  • Having it working in your infrastructure doesn't provide automation/orchestration of Disaster Recovery scenario. The same statement applies to Hyper-V Replica. You will need either powershell scripts or System Center Orchestrator to get automation options. 
  • Each Virtual Machines at the Recovery Site has to be powered on manually
  • You will need to reconnect each your VM to the correct network.
  • There might be necessity to Renumber IP addresses of some VMs - again, manually.
  • You will miss possibility to test your recovery plan during working hours since you cannot power on Replica VM if the original VM is still running and reachable.
  • There is no failback option in vSphere Replication. 

Well, it must be admitted that if you have all your steps of Disaster Recovery scenario well-documented and throughly tested you may deal with it just fine with probably 10 to 100 virtual machines. 
However, every single manual action significantly increases risk of human error, especially when every minute matters. Multiply all steps need to be taken by hundred/thousand VMs running in big enterprises  and you understand why it is not applicable. You can also take into the account the fact that there is not always a qualified person at the Recovery Site who can take care of all DR procedures properly.

Nevertheless, it looks like an adequate answer to Microsoft, considering that you can protect all your VMs at no cost if you own license starting from Essentials Plus Kit to Enterprise Plus Edition. An ideal solution for small companies seeking how to cut expenses on Disaster Recovery.

If you find this post useful please share it with any of the buttons below.

VMware vSphere 5.1 Clustering Deepdive is available now

Duncan Epping and Frank Denneman, two very probably the best VMware evangelists, have just released the new book covering every single bit of vSphere 5.1 Clustering.

I have bought and read both previous books of them (both in paper and in kindle format, for the simple reason that I can afford it :) and I can say that these books have significantly contributed into building my knowledge and clear understanding of HA and DRS cluster's mechanic and logic. I personally believe that HA/DRS functionality is a heart of all vSphere package and therefore the solid expertise of vSphere cluster is a must for anyone who sees his future in virtualization world.

So, choose your preferable format and enjoy reading.


vSphere 5.1 is released


Update:  The vSphere 5.1 and vCloud packages are available for downloading as of today.  Enjoy. I will have couple of sleepless nights for sure :) 

Being on vacation I have missed the release of my favourite product in IT world. I am trying to catch up now reading all "What's new" documents, but I definitely miss my lab to have some hands-on experience with all new fantastic features of vSphere 5.1.

Here are the links for all of the documents that can give you a brief overview of most interesting innovations in vSphere 5.1
There are probably some other VMware products which have had new versions, but I think the links above cover most of the important topics.

Tuesday 21 August 2012

Reclaiming Disk Space from Thin Disks

There are huge number of posts describing how to reclaim disk space and to decrease used storage for Thin Disks. However, very few pay attention or emphasize one precaution measure you have to take before proceeding.

Here is the whole story. There are situations when you temporarily COPY large amount of data on Thin Provisioned Disk. This action inflates disk's VMDK file on VMFS datastore. When you delete this data off the Thin Disk  the VMDK file size stays the same and that is because the data is not deleted, but simple marked to be overwritten. So you end up with situation when actual disk size in Virtual Machine is, let's say, 10 GB, but its VMDK occupies hundreds of GBs on VMFS Datastore.

Basically, this problem is resolved in two steps.
  1. You run SDELETE utility (link) on your drive to zeroize all unused disk space. This way you will let vSphere distinguish between actually used blocks of disks and unused disk space.
  2. Once it is over you need to start Storage vMotion of your Thin Disk. The main point here is that you have to move your Disk/VM to a VMFS Datastore with different Block Size. Under this condition only vSphere will invoke legacy datamover which doesn't copy zeroized data blocks. As usually, Duncan has the best explanation of the block size impact.
With VMFS-5 which has unified Block Size of 1 Mb this is not as easy as pie. First you will need to create a new VMFS-3 datastore with different block size and then it has to be upgraded to VMFS-5.

And here comes the  !Warning!
  • Pay Attention to the amount of Free Space on your VMFS Datastore before running SDELETE.
As I have said  SDELETE will zeroize all free space on you Thin Provisioned Disk. That means it will inflate VMDK file to its maximum provisioned space. So you have to make sure that SDELETE will not gobble all your free space on a VMFS datastore.

If you find this post useful please share it with any of the buttons below. 

Monday 20 August 2012

I am finally back

Well, I know for sure I am not getting vExpert status next year. Have been utterly busy with preparation for IELTS exam over the last 5 months. Just yesterday I have got the results I sought (Overall: 8.0).

I have been working with myriad of technologies this year, except VMware products :)  Therefore, I have got a bit rusty in a virtualization world and need a deep dive for at least a month into virtual products before I can get back to the level I was last year.

This week I am planning to write about HP Virtual Library System and not very positive experience with this product, though it could be just a result of using VLS together with Veritas Backup Exec.

Later on, I am going to share my experience with VMware Site Recovery Manager which I have planned and configured for our company last week.

And the last great piece of news is that now we have an authorized examination center in Kazakhstan where we can take VCAP 5 exams. I remember that even in Italy there was no such place where I could sit VCAP 4 exam last year and had to go to Switzerland.

See you soon guys.

PS I was really impressed by the fact that average number of visitors haven't decreased lately even though I haven't posted for so long. Thanks to you all for reading my blog.