Hyper-V offers fully-supported, application-aware, transactionally-consistent backups of any applications that have a VSS writer. These include Exchange, SQL Server, Oracle, SharePoint, and others. There is no need in Hyper-V to put an agent in any VM. Your applications will get properly backed up and they will know they’ve been backed up (thus clearing their logs) — without any agents. VMware, on the other hand, offers no such support — even though the functionality to do so has been available for seven years. And I think it’s time we talk about it.
BTW, if you’re not that familar with Microsoft VSS and all it allows you to do, you should read this blog post I wrote last week first.
This all started with a blog post from Scott Waterhouse of EMC that said that the only way to get application backups in VMware was to use an agent-based backup like Avamar. My immediate reaction was that Scott was off his rocker. He had no idea what he was talking about. (Wouldn’t be the first time I thought that, eh, Scott?) So he and I talked via email and then he called his people and I called my people. Along the way I had a Twitter epiphany which I blogged about.
There are three features that are very important here:
- Can Hyper-V/VMware get a consistent backup of a virtual volume within a VM (i.e. take a snapshot of the volume via VSS and then back up that snapshot)
- Can Hyper-V/VMware get a consistent backup of an application residing with a VM (i.e. Use VSS to quiesce the app, use VSS to take a snapshot of the volume(s) the app resides on via VSS and then back up that snapshot, then release the app)
- Can Hyper-V/VMware tell the application that is has been backed up so that it can clear its transaction logs (i.e. use VSS to tell it that it has been backed up)
Number 1 is table stakes. If you can’t do that, forget it. Number 2 is also important obviously. However, without Number 3, you’d need to run a backup agent just to tell the app to clear its logs. With that in mind, the following table explains what is possible in both products.
|Consistent backup of a volume||Yes||Yes||Yes||Yes|
|Consistent backup of applications||Yes||Yes||Yes||Yes|
|Applications aware of backups||Yes||Yes||Yes||No|
VMware obviously can back up a Windows volume via VSS. However, its support for applications is sadly lacking. (Update 11/30/10: They addressed the Windows 2008 issue.) It can only quiesce applications if they happen to be running on a Windows 2003 guest, and it It cannot tell any of the applications they have been backed up.
What this means is that anyone wishing to get proper backups of applications in Windows must run an agent of some kind in their guests in order to make this happen. Please note that this is not to say that you must perform guest-level backups. While those may like sound contradictory, statements this post should help clear up the fact that they’re not.
This means that any backup tools that are using only VMware’s infrastructure are going to have the same limitations. These include, but are not limited to:
- Almost all mainstream commercial backup applications are using the VCB and vSphere backup APIs and nothing else. If VMware isn’t quiescing an application and telling it [the app] it’s been backed up, then neither are they. This is one of the reasons why so many people continue to do guest-level backups of VMs to this day.
- VMware Data Recovery (of course) only uses VMware tools and can’t properly back up applications either.
- Vizioncore’s vRanger Pro also uses VMware’s VSS tools, according to this this forum post from last month. I reached out to them for comment and they did not reply. [Update: They have also addressed this issue.]
- PHD Virtual’s ESXpress didn’t appear to be using VSS at all. (Try searching for the phrase in their documentation.) However, I did find this forum post via a google search that talks about an undocumented feature for doing this. They also have not replied to my request for comment. [Update: They have addressed this issue.]
A few rays of hope
Another way to backup VMware has been using NetApp’s SnapManager for Virtual Infrastructure (SMVI), so I took a look at their website. But their product page for that product says that they work “in conjunction with VMware snapshots” which… call.. the VMware VSS requester. So I wasn’t feeling very good about that until I talked to NetApp. They confirmed that if you want to perform application-consistent snapshots, you want to use SMVI in conjunction with the appropriate SnapManager product (e.g. SnapManager for Exchange, SQL, or Oracle). These SnapManager products will place an agent in the guest, but the only purpose of the agent is to talk to Exchange and VSS to get it to do what it needs to do before it’s backed up. If you do this, NetApp says the products work together to “do the right thing.”
Veeam has a similar approach (which is now mirrored by their competitors). Their website says, “Unlike other vendors featuring limited VSS support, Veeam provides the most complete implementation of VSS support, equipping you for proper restore of VSS-aware applications (e.g. Active Directory, Exchange) from the created backups.” Hmm, this sounded interesting. I did a google search against their site and turned up several forum posts like this one where they’re using all the right terms (which I’ll cover later in this post). I contacted them and they clarified that they do use VMware’s VCB or vStorage API to do backups without sending data through the guest, but enhance VMware’s capabilities by placing small agents in the guests that they communicate with at the appropriate time to do the right thing.
So this is what I meant when I said earlier that putting an agent in a guest did not automatically translate into doing guest-level backups. Sometimes the agents are there just to coordinate things.
Another ray of hope is that, although they won’t say when it’s coming, my VMware contact told me that they are working on full support for VSS backups including telling the application that it’s been backed up.
And now I’m steamed
Do you have any idea how difficult it was to drag up this dirty little secret? It should have been as simple as reading a support matrix on VMware’s website. There is no such matrix for the vStorage API and you can only find the VCB version of it with a lot of digging. And even when you find that page it appears that they do the right thing in Windows 2003 (at least). It took a phone call to a very helpful VMware person to explain that this only means that they can quiesce the app; they do nothing to tell the app that it’s backed up.
My opinion is that this is just as important as which piece of hardware you need to buy to run VMware on, and it should be listed prominently in the VMware compatibility guides. This might prevent someone, for example, from upgrading from a version where they do support something to a version where they don’t support it.
The net/net of this is that Scott’s original post that started this needs a few corrections. He’s close, and he gets kudos for bringing the whole thing up. But you can get transactionally consistent backups without doing guest-level backups.
- As long as you’re still running Windows 2003, you You can get application-consistent backups using VCB or the vStorage API. It’s just that you’re going to need to figure out how to get them to truncate their logs.
- You can get fully-functional backups — and the apps will know they’re backed up — if you use NetApp’s SnapManager line or Veeam’s backup product (or any of the other products that now support this). They both do put an agent in the guest, but the guest is only for coordination and does no data transfer in the VM.
- And, of course, you could get all this wtih pretty much all backup products if you put your apps into Hyper-V.
----- Signature and Disclaimer -----
Written by W. Curtis Preston (@wcpreston). For those of you unfamiliar with my work, I've specialized in backup & recovery since 1993. I've written the O'Reilly books on backup and have worked with a number of native and commercial tools. I am now Chief Technical Evangelist at Druva, the leading provider of cloud-based data protection and data management tools for endpoints, infrastructure, and cloud applications. These posts reflect my own opinion and are not necessarily the opinion of my employer.