ESXi, backups, etc

Curio

Curio

Curio

Soldato
Joined
14 Mar 2004
Posts
8,040
Location
Brit in the USA
We're currently getting quotes from companies for server support and off-site backup. I'm having a hard time getting my head around how the VMs figure into the whole backup scenario.

We use ESXi 4.1 free, and as I understand it, the free version doesn't support backups? Only one company has identified this as a problem and say we'll need to spend thousands on the full version of ESXi to be able to backup the VMs. The other 2 companies have all-singing-all-dancing backup services, but no mention has been made of the VM side of things.

I've got so many acronyms and so much tech-babble BS flying around that I'm totally lost. The bottom line is, if our building gets hit by a meteor, will we be able to quickly get our VMs back on-line with ESXi free and off-site backup? Or do we need to spend a lot on a fully functional version of ESXi?

(please explain as if talking to a 2 year old with learning difficulties :p )

Thanks!
 
Have a look at William Lam's Ghettovcb scripts, by no means 'Enterprise ready' but they'll work.

It's been a while but from memory, the vStorage API for Data Protection (aka VADP) is read only when ESXi is unlicensed, thus it won't work.

Regarding the £1000s to upgrade, look at vSphere Essentials to see if it'll do a job for you, it runs to the £100s.
 
Straight off the bat, you need a clear distinction between whether you want Disaster Recovery (My building has been hit by a meteor and I need everything back up and running), backups (Some idiot in finance deleted our entire tax return last week and it's due tommorrow), or both.
There is some overlap, but your requirements need to be clear lest you end up with a solution that doesn't do what you wanted.


On with the techie stuff:
The guy who told you you need licensed ESXi was right, but his info is out of date.
With ESXi 5.1 the hotadd* functionality does not require any licensing - however, I don't believe that 2 of the big players in VM backups (vRanger and Veeam) have updated their software to take advantage of this yet, so for now a licensing constraint still stands.
(* Hotadd allows you to take a full image backup of a VM while it's running without impacting network traffic)


The best solution for you depends on an awful lot of things, like:

What are your backup routines right now?
What data are you backing up, Databases, fileshares?
How much data, and do you know it's rate of change?
How many servers?
Is it all VM's, or physical servers too?
How quickly do you want to be back up and running?
Do you have another site far enough away to avoid aforementioned meteor strike?
And the one everyone hates, what's your budget?


Now to try and answer your question:
Until backup utils are updated to take advantage of the relaxed licensing for hotadd (including the mentioned GhettoVCB scripts), ESXi needs to have at least an Enterprise license, and that is prohibitively expensive for a small company (€2.5K per CPU).

If you can't justify that cost, I wouldn't wait around for companies to catch up. I would buy some external hard disks and setup a rotation of overnight copies of your server data to them, and take them off site somewhere secure.
If the worst possible happens, at the very least you have several copies of your data saved.
 
Last edited:
OK, here is some more info. I'm not a server/network person so I might not be explaining this 100% correctly :p

We have 1 Dell R410 server which hosts 3 VMs. Our actual data is stored on an older Dell 1950, which is then replicated (ZFS) onto a second 1950. The two 1950's were our older servers that were re-utilized in this fashion when we got the R410. It was seen as a good option at the time because it meant getting more use from the 1950's. We then have a file-level backup to our current support company's data center using Crash Plan Pro.

Due to the disks in the 1950's not being very big, we have now run out of space. The plan (all 3 companies have essentially the same plan) is to migrate everything to the R410 which has 4TB (4x1TB drives in RAID 10) of storage that isn't even being used currently.

Company A (our current provider) is saying the problem with this is that we'll lose the redundancy of the ZFS replication. Although, I'm not sure where the RAID capability of the R410 comes in? Wouldn't RAID 10 give us that same level of backup as the ZFS setup?

Company B wants to install a dedicated backup server, which apparently does OS level backups for everything. This also then goes to their data center for off-site backup.

Company C I'm still waiting for the details from, but I get the impression it will be similar to B.
 
RAID isn't backup, if the controller goes, so does your data so company a are right in what they are saying
 
It depends, you could install backup software within the guest os and backup that way, or backup the vmdk files on the esxi host. In think backup exec can do specific VMware backups...
 
Uhtred said:
RAID isn't backup, if the controller goes, so does your data so company a are right in what they are saying
Click to expand...

Aren't there separate controllers? What's the point in having RAID 10 in a server if there's no redundancy? :confused:
 
Curio said:
Aren't there separate controllers? What's the point in having RAID 10 in a server if there's no redundancy? :confused:
Click to expand...

It's a fine line, but defining the terms I think is what's key. RAID provides redundancy, so when hard drives die etc it still functions. But if the server were to corrupt data in some way or something gets deleted, then RAID will not help with this. This is what a backup is for, a separate point in time copy of the server, ideally sufficiently isolated and removed from it's source to prevent negative causes affecting both live and backup data. I've seen a lot of people mix up the terms backup with redundancy and use them interoperability which causes confusion, even vendors..
 
Last edited:
Gammawolf said:
It's a fine line, but defining the terms I think is what's key. RAID provides redundancy, so when hard drives die etc it still functions. But if the server were to corrupt data in some way or something gets deleted, then RAID will not help with this. This is what a backup is for, a separate point in time copy of the server, ideally sufficiently isolated and removed from it's source to prevent negative causes affecting both live and backup data. I've seen a lot of people mix up the terms backup with redundancy and use them interoperability which causes confusion, even vendors..
Click to expand...

Thanks. How is our current ZFS replication different though?

Sorry, this is all very complicated to me and I'm having to get a crash-course in a short space of time!
 
To talk in terms of basic architecture and putting vm backups to one side for the moment it sounds like company b want to implement disk to cloud solution (great). Currently you have disk to disk (zfs replication).
RAID on its own would be disk to nothing.

Company c might come back with a disk to tape solution or even tape on its own.

Each solution has pros and cons, for example disk to disk is quick but might not help if the building burned down, disk to cloud would be quick and provide protection if the building were to burn down, but it would be more expensive. Tape might be cheaper and fairly quick but requires someone to manage the media.
 
You must log in or register to reply here.
Back
Top Bottom