Is Veeam Replication of Exchange supported?

[mcpcwld]

First post so please forgive me if this has been addressed before, I did have a look but couldn't find anything that answered my specific question.

In this Microsoft blog post https://blogs.technet.microsoft.com/rmi ... a-support/ it states that MS themselves do not support the replication of Exchange via Hyper-V Replication.

We're running Veeam B&R 9 and it's my understanding that the replication component leverages the Hyper-V Replication technology. So should we/can we use it to replicate Exchange (2013 in our case)?

[foggy]

Your understanding is not correct, Veeam B&R uses Hyper-V VM snapshot capabilities and has little in common with Hyper-V Replication.

[mcpcwld]

Thanks for replying so quickly foggy, and for clarifying the operation of Veeam B&R, which I'm relatively new to. You mention that B&R leverages Hyper-V VM snapshots, but Microsoft say they do not support the use of snapshots with a virtualised Exchange server - https://technet.microsoft.com/en-us/lib ... .150).aspx (see section "Requirements for hardware virtualization").

Can you tell me Veeam's support position on running a replication job of Exchange 2013? I'd like to be able to do this as it would help in a DR scenario, but the last thing I want to do is find myself with no support when I might really need it.

[foggy]

Microsoft does not support non-application aware backups of Exchange:

Supported Backup Technologies

Exchange 2013 supports only Exchange-aware, VSS-based backups. Exchange 2013 includes a plug-in for Windows Server Backup that enables you to make and restore VSS-based backups of Exchange data. To back up and restore Exchange 2013, you must use an Exchange-aware application that supports the VSS writer for Exchange 2013, such as Windows Server Backup (with the VSS plug-in), Microsoft System Center 2012 - Data Protection Manager, or a third-party Exchange-aware VSS-based application.

And this is right what Veeam B&R does - it uses application-aware VSS processing before creating a snapshot, so VM snapshots that are used for backup purposes are application-aware, which guarantees required application consistency on restore.

[Igor Lukic]

Hello,

first, I have to add a bit to the subject:

As far as I know, Exchange Server is not supported with Hyper-V Replica.
I couldn't find out why but if the Exchange Team says not supported, then it is not supported. Okay.


But how can Veeam do a replication of Exchange? This should also not be supported by MS.
I couldn't find any statement from Veeam if Veeam Replication will left Exchange in a supported status after a recovery.
Even with VSS Application Aware Processing.
Do you have any statements for this Scenario?
If possible, what kind of problems can happen?

Let's say, we have only one Exchange Server 2016 with all functions enabled (No DAG) installed on Hyper-V 2016.
Veeam Backups the first Hyper-V host with a backup to disk and to tape.
The Veeam Replication replicates to a second Hyper-V host.

Can Veeam help me with Veeam Replication to have a DR?
What is the difference with Hyper-V Replica, when enabling "Configuring additional Recovery points" and "Volume Shadow copy Service (VSS) frequency" ?

In my understanding. when doing any type of replication, after an activation of a recovered VM, the VM should be in a mode like pulling the power cord of a physical machine and restarting the machine again.
Maybe the database will come up, maybe not. If VSS works correct, the Exchange logs should be truncated. So maybe there is a dataloss between the last VSS Replica and there are not all Logfiles available.
Then a repair with one of this suggestions: https://support.microsoft.com/en-us/hel ... -and-event should work.

Do you have any deeper information available?


Thank you!

[nmdange]

Veeam replication is the same as doing a VSS backup and restore of the system, so it is always application consistent. I can't say whether it is officially supported by the Exchange team, but it does work differently from Hyper-V replica in that respect.

That said, is there a reason why you aren't running a DAG between these Hyper-V hosts? You would not have any issues of data loss whereas both Veeam and Hyper-V replication would lose anything after the last time the replication ran.

[DGrinev]

Hi Igor,

Please review this short discussion above as it contains the necessary information.
Do not hesitate to ask additional questions. Thanks!

[Igor Lukic]

Hi nmdange
Hi Dmitry Grinev

thank you both for answering.

That said, is there a reason why you aren't running a DAG between these Hyper-V hosts? You would not have any issues of data loss whereas both Veeam and Hyper-V replication would lose anything after the last time the replication ran.

I had a discussion with some customers. To be clear: I don't know the excact configuration, this is only what they told me:

They are all currently using VMWare Hosts with one Exchange Server, no DAG, on a Fibre Channel SAN with two Disksystems which are replicating to each other.
So if a VMWare host brake, they restarted the Exchange Server and everything worked fine again.
In a situation, where one of those Disksystems failed, they could work with Exchange after a restart of the machine.

Specialy one customer was pretty direct in his question:
Because of a new admin, they want to switch over to hyper-v because in his opinion the current setup is oversized for this company.
He is thinking, that two Hyper-V Servers with direct installed storage in the box should be enough. He doesn't want a DAG. (He think it is simpler to have only one Exchange)
All incoming and outgoing mails are saved in an external mailarchiv. So there should be no "real" dataloss because this is an external box.

So where is the difference between Hyper-V Replica and a Veeam Replication with a snapshot?

In the discussion he started with what could happen if:
- Hyper-V Replication is activated and replication is done every 15 min. (NO VSS active)
In his opinion, the host with the replica should start the guest. hyper-V will integrate the last known changes in his vhdx file.
This vhdx will be used to start Windows and Exchange. Worst case is that in the last 15 min a huge amount of mails where send. Those mails are now not available. Exchange "should" be running.

When "enabling "Configuring additional Recovery points" and "Volume Shadow copy Service (VSS) frequency" the Exchange logfiles should be truncated, so the database should have at least the last hour of mails.
If not, he has to bring the database up againt
And as "foggy" said: "Microsoft does not support non-application aware backups of Exchange:"

but it does work differently from Hyper-V replica in that respect.

Where can I find information what the differences are?


So I told him about Replication with Veeam.

In this thread:
microsoft-hyper-v-f25/hyper-v-2012-r2-r ... 16374.html
the main reason to use Veeam is reduced disk I/O. (In his configuration, he has a lot of SSDs, so he thinks that this wouldn't be a problem.)
The other points are not interesting for him.

I would like to convince my customer to do it the "right" way.
But I need more arguments, the best would be a "killer" argument.

Thank you.

[nmdange]

If you did a planned failover where you shutdown the source VM and replicated all changes, then it would be safe. However, in the event of an unplanned failover, Hyper-V replica is crash-consistent, which the Exchange team explicitly does not support. Veeam replication would be application consistent so it would be ok. However, in both cases, if the source server dies completely, you will lose all the mail sent since the last replication interval. A DAG will ensure nothing is lost, since the logs are copied more frequently and the hub transport safety net catches everything else.

I've seen many organizations try to "out-think" the Exchange team by putting Exchange on VMWare with SANs, doing SAN replication, failover, etc. and then run into problems. Even if it works most of the time, it can never beat the best-practices design of a 4-node DAG spread across two datacenter (see https://blogs.technet.microsoft.com/exc ... hitecture/). DAGs are not hard to set up and Exchange by itself is going to be better at HA than trying to rely on something else. Honestly, if an organization is not large enough to justify the 4-node DAG design in the preferred architecture, then they are better off moving to Office 365. I realize that is a bit of a different discussion though

[DivideByZero]

Hi,

Is there a guide anywhere for doing a planned failover of a non-DAG Exchange 2010 box? Or any databasey type things?

The last time we attempted this we encountered some DB corruption, so I would like to know if there is a guide out there to make sure we are not missing thing obvious? I have seen lots of things about backing up and replicating DAGs, but this is just a single server environment.

Thanks

/0

[PTide]

Hi,

Have you seen the corresponding section of the User Guide already? If everything is configured properly and you still get corrupted DB then please contact our support team directly so they can take a closer look. kindly post your case ID when you have one.

Thanks you

[DivideByZero]

I had read that, but as I'm not the Veeam admin (just Exchange) it doesn't mean a lot to me. It says you can "Process transaction logs with this job" or "Perform copy only", but would we expect a dead DB on the other side with either of these options? I wouldn't have thought so.

Something else I was thinking is placing Exchange 2010 in to maintenance mode and shutting the services down BEFORE we do the failover. Is this supported? Can't find any documentation to say if this is supported or a bad idea.

We can certainly log an incident if we run in to this issue again, but knowing what to do to avoid an issue would be a lot better!

[Regnor]

During a planned failover the Source VM is shutdown during the process, the delta changes are getting replicated and afterwards the target/replica VM is powered on.
So there's no need to manually stop services or the VM itself, as the process creates a consistent replica.
https://helpcenter.veeam.com/docs/backu ... tml?ver=95

[PTide]

James,

To ensure that the database in the replica is consistent and alive you have to select "Require successful processing" option. Is it selected?

Thanks

[Alan_ORiordan]

So what I gather from this thread is that it is supported due to Veeam having application-aware processing? When Exchange is failed over for testing are there any special changes made to AD that would cause an issue when the failover is undone ie all changes disregarded and not synced back to live?

I am slightly disappointed/concerned that there are not many articles/videos on this being demonstrated like people are afraid to do it?

I have used Veeam Failover to great effect on application, SQL and Linux machines but never an Exchange server, yet...

If we talk about a simple scenario where:

LIVE Site

Has two DC's
One Exchange 2013 or 2016 server (Standalone/basic config)

DR Site

Has it's own DC on a different subnet

During a planned failover with an application-aware replica the Exchange_replica should power on and Re-IP and network map itself.
It will register its Re-IP'd address in DNS and will eventually update the LIVE Site DNS (if failover lasts long enough) and the DC's at the live site are left online during testing
If we have split-brain DNS then autodiscover and internal DNS records will not update only the HOST A record will? Do people update these manually during testing?
If we want to undo failover then we should ensure that no email is delivered through the DR site firewall to avoid missing emails

Are there any other considerations to provide a failed over Exchange Server?

What is the best way to test the Exchange_replica, in its own failover plan for example?

Thanks