Veeam Replication Design Recommendations

[oldmantime]

Anyone have best practice recommendations concerning a veeam HQ and DR setup.

Which site should veeam be hosted/installed to clone virtual machines more efficiently?

Is is possible to veeam the veeam server from DR to HQ and then bring it online with the most current database?

[Gostev]

Assuming that by "cloning" you meant replication - this really depends on scenario, environment and ESX versions. In some cases, you will want to install Veeam Backup in source site, in other - in target. Please tell me more about your situation?

Generally you cannot have Veeam Backup server backing up or replicating itself, what happens is in such setup is this: Veeam Backup would freeze itself during the snapshot creation and will not be able to complete the required operations. But of course you can always backup Veeam Backup VM with another Veeam Backup install. Alternatively, if you use remote SQL, you should replicate SQL VM to DR location. In this case, DR procedure will involve installing Veeam Backup in the target site and pointing it to the existing (replica) database with all configuration settings. Veeam Backup install with existing remote DB takes just 1-2 minutes if you have server VM provisioned and ready to go.

[oldmantime]

The goal is to replicate physical and virtual machines from HQ to DR on a predetermined schedule. The DR site contains (4) ESX 4.0 servers with a virtual machine running the veeam software. HQ contains (3) ESX 3.5 servers and no veeam is currently deployed here.

I want to deploy the most effective solution and I'm not quite sure where to place the veeam server. Is it better to place it out in DR?

[Gostev]

For your scenario, Veeam server absolutely must be in HQ so that only changed blocks are transferred over WAN. ESX 3.5 does not support changed block tracking, so full VM will be retrieved from source ESX storage to Veeam server each replication cycle, then changed blocks will be determined, and only they will be sent over WAN to the target host.

[cregan]

This topic is very interesting because I am asking myself the same question.

Currently my two ESX4 servers are in the same place.
I have two san. SAN1 is in the primary site and contain all my VM.
In my DR site I have SAN2 that contain all my replica and a physical Veeam/Vcenter server. DR and primary site are connected via 1gb fiber link.
I don't have an ESX server in DR site.

I was thinking of removing one ESX4 server from my primary site and putting it in my DR site connected to my SAN2 with some VM on SAN2, then replicating the VM to my Primary site on SAN1 and vice-versa.
In that case I can't figure where to put the Veeam server and the Vcenter server. I was thinking a third site.
The pro is SAN2 is used and it remove some load off SAN1.
The cons, It's complicated, I need to move a bunch of VM to SAN2.

Second option was to buy a whitebox server, install ESX4 on it and leaving it on the DR site to start the replica in case of disaster.
That solution is easy but the con is my SAN2 is only being use to store the replica. In that case I would put veeam and vcenter in DR site.

Also, if veeam DB is on a remote SQL server on primary site and that server crash, veeam can't restore that SQL server because it can't connect to the sql database.

What seems the best solution from your perspective ? Maybe you have some recommendations that I didn't think of...

[Gostev]

There are different pros and cons of installing Veeam Backup in source or target site... cannot think of any pros of installing it in 3rd site though...

Source site
PROS
• Only changed blocks are transferred over WAN (whether or not you use changed block tracking).
• Network traffic compression.
CONS
• If source site goes down along with Veeam Backup server, you will only be able to failover to the latest state by manually starting replica in vSphere Client.

Target site
PROS
• Only changed blocks are transferred over WAN due to the ESX4 changed block tracking.
• If the source site goes down, Veeam Backup will still be alive with its config DB, and you will be able to failover VM to both the current and to the earlier state via UI.
CONS
• You get no network traffic compression.

[greathound]

I too am looking for some recommendations. Our environment consists of a Head Office vCenter with our ESX hosts at branch offices across a WAN link. I am interested in D/R and backup options at each site which houses 2 ESX hosts (HA) and a SAN.

I'd like to know what the design might look like as far as placement of the Backup Console and the Enterprise manager. Do we put both on the vCenter server? Separate server? Backup console at each Branch Office?
SQL placement?

Thanks

[Gostev]

Gary, in your case I would go with one Veeam Backup server in each site, so that backups are performed locally. Since you have few VMs (only 2 hosts) in each site, you can install Veeam Backup on vCenter VM and use local SQL Express (unless you already have "full" SQL in each site), this should be adequate. You only need one Enteprise Manager installation, and you want to put it in HQ office.

[skayser]

Generally you cannot have Veeam Backup server backing up or replicating itself, what happens is in such setup is this: Veeam Backup would freeze itself during the snapshot creation and will not be able to complete the required operations.

Partly hi-jacking this thread for what Gostev said above regarding the self-backup of the VBR VM. The release notes for 4.1.1 say that it's not possible to backup VDR itself using the Virtual Appliance mode ...

Virtualized Veeam Backup server cannot backup, replicate or copy itself with jobs using vStorage API “Virtual Appliance” mode.

... which sounds to me as if it would be possible in other modes. How are the release notes to be read? VDR able to backup itself or not?

[Vitaliy S.]

Sebastian,

Yes, you may use Veeam Backup to backup itself in all the modes except Virtual Appliance mode. Please note that you need to make sure Veeam VSS and VMware Tools queiscence are turned off. But I wouldn't recommend doing that if your SQL Servers is active due to being used by some other apps.

[Gostev]

What is the best way of protecting the Veeam Backup server?

[nicolas.gras]

Hi,
I need a clarification regarding licensing :

If i have a veeam Backup server on HQ site for local VM backup and a other Veeam Backup Server On DR Setup to manage replication (Mainly to failover to any version of a replica If source site goes down);
I have to buy other Veeam licenses for DR site ? (Source servers are the same in both cases...)

Thank's

[Vitaliy S.]

Nicolas, you may use the same license file for your main and off-site backup servers provided that the license file covers the same source hosts and this number is not exceeded.

[nicolas.gras]

Thank's Vitaly for your quick response. (In addition the content of the response I like also ).

A other question about a veeam HQ and DR setup scénario :

If i installing Veeam Backup in target site (always, to failover to any version of a replica If source site goes down). It is possible to Setting up the initial replication to external storage (USB Disk) in source site ?

Thank's

[Vitaliy S.]

Unfortunately, no, as initial replication is only possible with Veeam Backup server installed in the source site.

[nickf]

Sorry to resurrect this one but could someone clarify the Con listed against Pull replication that there is no network compression. Does this mean you can't use a WAN optimiser or is there network compression built into Veeam that will only work on push? We're looking at switching from Push to Pull replication to make it easier to deal with a mid-replication DR event but we're also only just coping with the amount of data we're trying to replicate (not currently using WAN optimsers) so wan't to be sure I don't make things worse by switching to Pull (the hosts are ESX 4.1 so the changed block tracking shouldn't be an issue).

[Bunce]

the vbr server needs to see the traffic before it can compress it so it would have to come across the wan first - hence the Con.

no reason not to use a wan optimizer though, in fact it would be even more critical in this case.

[st0623]

Guys, HyperIP is a Veeam Partner for WAN Optimization/compression. Download an eval copy of the virtual appliance here, http://www.netex.com/eval. We'll let you use HyperIP for 30 days for free at any speed. Reach out if you have any questions: steve.thompson@netex.com.

A quick note. WAN compression usually tops out at around 200Mbs. 2:1 ratio is 400Mbs, 3:1 is 600Mbs of application throughput, etc and so on. Veeam Replication should drive these kind of speeds over a 1Gbs link. Also, our compression in HyperIP can be set up for push or pull.

[ZackAngelo]

Pull is superior if you have a WAN optimizer. The reason is two-fold: 1) The WAN optimizer will perform the compression inline anyway and 2) the WAN optimizer will be unable to take redundant data off the network if it is compressed first.

To give you a basic idea, we've been testing Veeam replication on one of our Exchange 2010 CAS servers and it took 13hr 43min to do an initial replication of 32GB over a T1 WAN-optimized (Riverbed Steelheads) network. A follow-up replication six hours later took 11min.

[fbroussey]

HEllo All,

I'm sorry if my question is a newbie one

In all you posts, you speak about DR site accross WAN Link but what about DR site on LAN?
I mean my customer has a second room in another building with a Fibre between the 2 buildings (1gb/s).
In the "HQ" building he has 2 esx 4.1 with a SAN and we want to replicate his vm's on the "DR" Building (1 ESX4.1 with local storage). So my first attention is to put the Veeam DR VM on the "DR" site in order to have it available in case of a big failure on the HQ's SAN . But is it the good solution ?

Thanks for your design help

Regards,

[Gostev]

Hi, for replication over LAN, Veeam server placement does really affect performance - so you can do it in whatever way you prefer. Thanks.

[jamesr]

HEllo All,

I'm sorry if my question is a newbie one

In all you posts, you speak about DR site accross WAN Link but what about DR site on LAN?
I mean my customer has a second room in another building with a Fibre between the 2 buildings (1gb/s).
In the "HQ" building he has 2 esx 4.1 with a SAN and we want to replicate his vm's on the "DR" Building (1 ESX4.1 with local storage). So my first attention is to put the Veeam DR VM on the "DR" site in order to have it available in case of a big failure on the HQ's SAN . But is it the good solution ?

Thanks for your design help

Regards,

This is a similar question as I had around the placement of backup servers. We also have two server rooms within a campus on the same LAN with a SAN in each sharing a stretched storage fabric. I'm trying to decide between-

a) locate backup server in room A backing up VMs from SAN in room A, backup LUN target on SAN in room B

or

b) locate backup server in room B backing up VMs from SAN in room A, backup LUN target on SAN in room B

[bc07]

Check out also this thread: replication and ESXi

[stu0132]

Hi,

I am looking for some advice. I have two sites seperated by a national ethernet 100mbit. We are replicating via veeam to the dr site and are only getting about 20-30mbits per second line saturation of Veeam.
We have proven the line as when we run a FTP copy we can see the line saturate. When we run 5 replication jobs at once we hit 60mbit of bandwidth. Should i be getting higher throughput ?

There are 5 VM hosts at hte main site and 2 at the dr site. Veeam is at the main site. Any pointers would be helpfull.

[Vitaliy S.]

Where are you replicating to? Is it ESXi or ESX host?

[hagepat]

Hi there,

We have 1 ESXi 4.1 host in HQ and 1 ESXi 4.1 host in Remote office. All hosts will replicate with eachother. So all vm's on ESXi host HQ will replicate over 10MB Fibre to ESXi 4.1 host Remote Office and vice-versa. In both sites I have an Veeam Backup and replication server. Because ther's a lot of data to replicate I need initial seeding. I think best way to do is create veeam replica job on HQ Veeam (push) to ESXi 4.1 host Remote Office, and vice versa off course. This is because seeding does not work with pull replication. Can you please confirm?

[Vitaliy S.]

Yes, initial seeding is only possible with backup server located on the source side, however our version 6 will make this requirement go away, meaning that initial seeding would be possible from both source and target sites.

Please also be aware that push replication is not a very efficient way (in terms of job performance) to run replication jobs targeted to ESXi hosts, but again this is also going to be changed in version 6.

Thanks!

[mdornfeld]

So based on what Vitaliy just said, in v6 is there an updated Pros/Cons chart of whether to do push/pull replication?

Also, we still need to backup to tape as well. Is there any advantage of having the tape drive at the DR site vs HQ? Should it be at HQ since it's a completely separate Veeam job and you want the backup near to tape, or is there anyway to be able to do a tape backup based on the replication job?

Thank you.
Matt

[Gostev]

Hi Matt, v6 has completely new replication design without notions of push or pull replication.
As for tape, I guess you would want your tape drive near the data you want to backup...
Thanks!

[mdornfeld]

Thanks Anton, but is there a way to do a Veeam backup job off of the replicated Veeam data, or does the Veeam backup job always have to run off the source VM?
Thank you!