Hypervisor preference - costs included
-
@tim_g said in Hypervisor preference - costs included:
@scottalanmiller So what exactly are the downsides to hypervisor / agentless VM backups that agents on everything resolves?
Mostly it's human stuff. Agentless backups make people get really lazy about how they back up. It's not that agentless can't do the job, but no agents out there work broadly enough to be reliable enough to use for any company that I deal with in the real world. Every company I know has at least one workload that requires agents, so we are stuck with them already. It's a nice idea and if you fit perfectly into the really niche mold of 100% supported OSes, Hypervisors, and applications that your agentless solution supports, then great, it might be for you. But the degree to which a company has to design everything around its backups and continue to do so for forever is pretty extreme. One casually added workload that doesn't fit that exact mold and either you just accept the crash consistency risk, or you have to move to agents (or some other alternative.)
Agentless makes people think that they don't have to know what all is being backed up and manage it correctly. It is treated as a panacea. It has a place, for sure but basically I'd never run agentless without an agent, as well. At best, it's an add on, not a replacement.
-
@tim_g said in Hypervisor preference - costs included:
@scottalanmiller said in Hypervisor preference - costs included:
@tim_g said in Hypervisor preference - costs included:
And if I need to restore a VM or file it's so much faster.
If you need to restore an entire VM exactly as is, to exactly where it was, if your backup retention location is local. That's a very specific, but reasonably common, circumstance. But for most companies, that's not the main restore scenario. It's pretty rare that I am restoring an entire VM "as is".
That's one aspect of it, right. Most need file restore, which you can do too. So why not agentless if it does both and everything is much faster?
Because it doesn't do both. No Agentless on the market handles any broad application set. Veeam is the best, and is very limited. Exchange and SQL Server are the only big apps that it supports.
-
Now if we are assuming that our key data is backed up via something automated that we make ourselves. And we are only considering agent vs. agentless for our fileservers, then by all means, most agentless solutions handle this very gracefully and that's great. Use it for that.
But it generally means that we are only talking about a fileserver when we start talking about backing up in this way. That's normally a pretty trivial amount of the overall workloads. Restoring a single file is really only common with file servers, nothing else. Maybe desktops, but agentless doesn't apply to normal desktops (maybe VDI, sure.)
Outside of that one use case, where either works fine under normal circumstances, most all workloads either require something special or find agentless to not really be advantageous.
-
What are most people using as a backup system for KVM? I've mostly just used StorageCraft and Veeam for Hyper-V and VMware.
-
@scottalanmiller said in Hypervisor preference - costs included:
@tim_g said in Hypervisor preference - costs included:
@scottalanmiller said in Hypervisor preference - costs included:
@tim_g said in Hypervisor preference - costs included:
And if I need to restore a VM or file it's so much faster.
If you need to restore an entire VM exactly as is, to exactly where it was, if your backup retention location is local. That's a very specific, but reasonably common, circumstance. But for most companies, that's not the main restore scenario. It's pretty rare that I am restoring an entire VM "as is".
That's one aspect of it, right. Most need file restore, which you can do too. So why not agentless if it does both and everything is much faster?
Because it doesn't do both. No Agentless on the market handles any broad application set. Veeam is the best, and is very limited. Exchange and SQL Server are the only big apps that it supports.
Yes, it does do both. I said agentless backups do both VM level and file level back up and restore.
You are now bringing in a 3rd apsect... application level backup and restore.
That's a different can of worms, but even agentless backups do any application level backup that has or supports vss.
-
@bbigford said in Hypervisor preference - costs included:
What are most people using as a backup system for KVM? I've mostly just used StorageCraft and Veeam for Hyper-V and VMware.
StorageCraft is agent based so works exactly the same on KVM. Veeam has optional agents, so works just fine there. Our big KVM is Scale HC3, which has built in backups.
-
@tim_g said in Hypervisor preference - costs included:
You are now bringing in a 3rd apsect... application level backup and restore.
This is the actual purpose of all backups. To get apps back to where they were. They do OS and other things only as artefacts of getting back to the data and systgems working.
-
@tim_g said in Hypervisor preference - costs included:
That's a different can of worms, but even agentless backups do any application level backup that has or supports vss.
That's true, but loads of things do not.
-
@tim_g said in Hypervisor preference - costs included:
Yes, it does do both. I said agentless backups do both VM level and file level back up and restore.
Right, but do you need either of those things? I totally get the "just restore the whole thing from an old snapshot" method, it is super quick and easy. But it's also unnecessary. File servers are an "app" that uses files, so that's an app specific situation. Other than that, it's all stateful data in unknown states that we have to deal with. So yea, agentless does the parts we are least concerned about. I have unlimited options for those parts. It's the apps that matter and are hard. And since I need agents 99% of the time because of the workloads, the agentless is just "more to manage". Sometimes worth it, but always extra effort.
-
@scottalanmiller said in Hypervisor preference - costs included:
@tim_g said in Hypervisor preference - costs included:
Yes, it does do both. I said agentless backups do both VM level and file level back up and restore.
Right, but do you need either of those things? I totally get the "just restore the whole thing from an old snapshot" method, it is super quick and easy. But it's also unnecessary. File servers are an "app" that uses files, so that's an app specific situation. Other than that, it's all stateful data in unknown states that we have to deal with. So yea, agentless does the parts we are least concerned about. I have unlimited options for those parts. It's the apps that matter and are hard. And since I need agents 99% of the time because of the workloads, the agentless is just "more to manage". Sometimes worth it, but always extra effort.
Not exactly.
We have file servers.
We have web servers.
We have Apps... that are stateless, but connect to a database that is stateful, but does it's own backups (MS SQL).
We have web servers that use MySQL and such, also do their own database backups.
I don't believe we have any stateful apps, besides databases that take care of themselves anyways.I guess I'm living in one of your rare cases Agentless backups make sense, because we have our VM level, file level, and app level protected and restorable.
-
@scottalanmiller said in Hypervisor preference - costs included:
Now if we are assuming that our key data is backed up via something automated that we make ourselves. And we are only considering agent vs. agentless for our fileservers, then by all means, most agentless solutions handle this very gracefully and that's great. Use it for that.
But it generally means that we are only talking about a fileserver when we start talking about backing up in this way. That's normally a pretty trivial amount of the overall workloads. Restoring a single file is really only common with file servers, nothing else. Maybe desktops, but agentless doesn't apply to normal desktops (maybe VDI, sure.)
Outside of that one use case, where either works fine under normal circumstances, most all workloads either require something special or find agentless to not really be advantageous.
I just seen this post. Yeah.
For the stateful servers like MS SQL, web databases, AD, etc, they are set up to do addional backup besides the Agentless hypervisor level backups.
-
@tim_g said in Hypervisor preference - costs included:
@scottalanmiller said in Hypervisor preference - costs included:
Now if we are assuming that our key data is backed up via something automated that we make ourselves. And we are only considering agent vs. agentless for our fileservers, then by all means, most agentless solutions handle this very gracefully and that's great. Use it for that.
But it generally means that we are only talking about a fileserver when we start talking about backing up in this way. That's normally a pretty trivial amount of the overall workloads. Restoring a single file is really only common with file servers, nothing else. Maybe desktops, but agentless doesn't apply to normal desktops (maybe VDI, sure.)
Outside of that one use case, where either works fine under normal circumstances, most all workloads either require something special or find agentless to not really be advantageous.
I just seen this post. Yeah.
For the stateful servers like MS SQL, web databases, AD, etc, they are set up to do addional backup besides the Agentless hypervisor level backups.
Oh okay, that's how we are. For everything that isn't stateless, we can build it from scratch without needing to do a restore, so no need to backup at all
-
@tim_g said in Hypervisor preference - costs included:
We have web servers that use MySQL and such, also do their own database backups.
I consider this an agent. Maybe that's where we have confusion. You are using what I would consider an agent, even if it just does a "fake" backup locally that the agentless then picks up.
-
@scottalanmiller said in Hypervisor preference - costs included:
@tim_g said in Hypervisor preference - costs included:
I for one don't have the time to deal with hundreds of agents...
What time do they take? If they are part of your build and automation process, they should take essentially no more time than agentless. It's all automated, right?
Just out of curiosity: if you have not a central management system (if you like it call it a single pane of plastic - not much glass nowdays on monitors) can you programmatically setup/modify/reschedule agent based backups with -just to say- veeam agent free?!
Never tried it.
-
Also one thing to consider is that some hypervisor level backup solutions will cost in the 500€-750€ ballpark. A single paid version of veeam agent server (which does CBT) costs 130 per guest! And I cite veeam just because it is widespread.
Even with 10 machines this is a bit of a cost IMHO. In my previous work I've spent 500€ for a backup solution which managed 12 VMs - with all the fancy stuff: CBT, dedup, incremental, blablabla and yes: a single pane of plastic.
I do not think that costs are to be kept out of the equation here.
-
@matteo-nunziati said in Hypervisor preference - costs included:
@scottalanmiller said in Hypervisor preference - costs included:
@tim_g said in Hypervisor preference - costs included:
I for one don't have the time to deal with hundreds of agents...
What time do they take? If they are part of your build and automation process, they should take essentially no more time than agentless. It's all automated, right?
Just out of curiosity: if you have not a central management system (if you like it call it a single pane of plastic - not much glass nowdays on monitors) can you programmatically setup/modify/reschedule agent based backups with -just to say- veeam agent free?!
Never tried it.
With Veeam? I believe so, but I've not tried. I think that their Linux agent at least uses config text files, though.
-
@matteo-nunziati said in Hypervisor preference - costs included:
Also one thing to consider is that some hypervisor level backup solutions will cost in the 500€-750€ ballpark. A single paid version of veeam agent server (which does CBT) costs 130 per guest! And I cite veeam just because it is widespread.
Even with 10 machines this is a bit of a cost IMHO. In my previous work I've spent 500€ for a backup solution which managed 12 VMs - with all the fancy stuff: CBT, dedup, incremental, blablabla and yes: a single pane of plastic.
I do not think that costs are to be kept out of the equation here.
Very true. But in most cases I look at, we don't use an agent of that nature. Veeam's agent is great, but I'd see it as more special case.
