I am at a new company and we are engaged in a very hot discussion about virtualization and the value of doing regular VM backups. The long term employees here have not real Vmware/Virtual experience but have read about it. Also, this is a small company and that addresses some of the reasons for the discussion.
Everyone likes the idea of virtualizing our servers (consolidation). But, they don't see a value in doing regular VM backups...but, rather they believe we shoud setup our P2Vs and even new VMs so that the data is stored in Raw Disk format and never as virtual disks. Then, they feel that doing more than a single backup of the VM is all that is necessary since without the data, the VM pretty much remains unchanged and therefore there is no reason to do regular VM backups. This way, they don't need to have a separate large storage server to keep the VM backups and we can just stay with local storage or external drives to backup any other physical drives....or , save the Raw disks on the external hard drives in separate disks and move them offsite.
Personally, I favor keeping as many data disks as vmdk so long as they aren't LARGE....some data need to be on Raw disks....ie SQL data, Exchange datafiles, for example. But, for the most part most VM disks are not going to be large and we can keep them with the VM and thus have the ability to restore them with the VM, if required or...if we use tools like VRangerPro....we can even do file recoveries from the backed up VMs. Moreover, there is a tremendous value in that it saves a great deal of time when dealing with disaster recoveries or large scale failures.
Personally, I don't like to use Raw disks unless absolutely necessary. But, that's just me.
In any event, I feel very uncomfortable about not doing regular VM backups
Anyway, I thought I'd ask the community to see what I am missing in this argument. Or, am I overly concerned for nothing.
I most cases RDM are used in the wrong way... There isn't a real benefit in performance. And in vSphere 4.x also a RDM has the same vmdk size limit (max 2 TB).
RDM must be used in guest cluster environment... Then may be used in some other cases... but only same and if possible in virtual mode (so you still have the same advantages of vmdk, like snapshot).
Having an image level backup that can also do file level will allow you roll back entire machine states and depending on which software you choose - do an instant restore from backup - I've lost count of the number of times misguiding application teams have believed that a system state restore is like having a bare metal backup.
keeping your virtual disks on RDM is going to make your storage management very tricky and could potentially limit the number of VM's you can host ( though this may not be a problem at the scale you mention )
In terms of RDM vs VMDK for SQL - if you plan your datastores correctly and put them on the same grade of disk as you would put your RDM's then there is no reason for them to be be any slower than using RDM , and with a lot less headache. As Andre says - the only time you would really need them is in MS Clusters or if you required specific Storage based features.
Moreover, there is a tremendous value in that it saves a great deal of time when dealing with disaster recoveries or large scale failures.
To me this statement of yours sums it up best. The idea that the VM only changes once in while, it is just the data, is an approach that sounds ripe for problems.
I might also add that really it's about Protection. Protection means different things, and different layers. While it might be an exercise in scemantics, I find that using the term 'backups' with those who are used to traditional environments just don't get it. I had to do a healthy amount of explaining why in many cases, protecting the systems that serve up the data is as important as protecting the data itself. I demonstrated along the way that this is the aspect that traditional protection of physical systems typically falls short.. Anyone who has had to rebuild an a solution from nothing but databases, etc. knows what I'm talking about here. CRM's, CMS, Exchange, etc. all provide a level of business logic that has an indelible tie to the systems that serve them up (registry entires, GUIDs, etc. Sure, you might, and probably should have the systems at a frequency that is a bit different than the respective data volumes, but ultimately you are looking at how best to protect the systems that provide the solutions for your business.
Good luck to you.
I had a recent failure of a raid array that took out not only 1 of my VMFS' but also an RDM. If I did not have a current backup of the entire VM I could not have restored my data to temporary storage as quickly as I did. I was down for less than 1 hour for the major VMs (all of which were fairly small) but for the RDM which was larger I was unavailable for considerably longer, but if I lost the data I would have had a complete failure.
Backups work and are there to help in such unforseen disasters.
There are hundreds of cases of such failures and quick recoveries, but there are equally if not more cases of failure to backup and crucial data was loss which would lead to perhaps a company failing, crucial designs being lost, or even failure to get the funds necessary to continue working.
Some backups are once a month, others are incremental, others are less often or more often depending on the data and such. Disk to Disk to Tape (or BluRay media) seems to be very very common in smaller businesses. Sometimes it is disk to disk to disk... If you are thinking about using the cloud, then I suggest encrypting data first.
Make yourself a rockstar, backup! And test those backups!
If they cannot see the benefit of regular backups, fail the system and see how long it takes to recover. Of course, you better have the okay of the people in charge first.
Edward L. Haletky
Communities Moderator, VMware vExpert,
I always thought that being able to do a VM image level backup was one of the best things that virtualization did for you. I have done some DR tests where all we had to do was restore the backup and power the VMs back on, can't get better then that. As stated today with current backup solutions there are so many more other benefits that are provided like File Level Restores. But even without the more recent bells and whistles in the virtual backup arena having that image backup of the entire VM ( data included) is key.