VSS question regarding backup/replication of SQL, DC...

[geordi]

Hi there

At the moment we are using VBR5 (latest patch level) "only" for creating daily backups during the night. The next step is to use the replication part on an hourly base during the day and one backup job during the night.
I have tested that and the machines are freezing for some seconds which is resulting in application errors.

The machines we want to replicate are Win2008R2 with actual patches and VM Tools (ESXi4 and ESXi5 with all available patches). We are using a shared storage connected via NFS. We want to replicate our Exchange, SQL and DC with AD-integrated DNS.
I have turned on the application awareness in the job settings from VBR. My understanding is/was that this would trigger the VSS from Microsoft and there is no relevant impact on the system during the replication.

Do I misunderstood the whole part completely or where is my fault in that scenario?

Thanks in advance,
Michael

[dellock6]

VSS goal is to allow for application-aware snapshots, freezing the application and truncating logs. It has nothing to do with performances, BUT if your storage has limited amount of IOPS, it could be a problem. Usually VSS freezing is a matter of fraction of seconds, if the freeze takes longer, you need to discover why.
What kind of storage do you use, and how many IOPS are produced by those VMs you want to freeze?

[geordi]

Good question, I have no idea how many IOPS the machines are generating ... what is the best way to measure that? esxtop?

The storage is a Synology RS3411XS with an additional shelve ... 22x WD Raptor 10k 600GB RAID-10.

But we are planning to buy an Eternus DX80S2 in the next time

[Gostev]

VSS freeze behavior differs from application to application, basically it is up to application's developers to decide how the application should prepare itself for hot backup. You would assume that, for example, in case of database application it should stop accepting new connections and queries for a short period of time, and commit any existing, on-going transactions as a part of the freeze. So indeed, there might be a short time period when such application will not available to client request.

Now, poorly written client applications may report issue with that immediately. However, more enterprise-ready apps will not start throwing alarms right away, because network glitches and server application overloads are quite common anyway. Any well-written client app acknowledges this, so it would first perform a few retries separated by timeouts, and only after that concludes that application is unavailable, and logs the corresponding event.

Example of "proper" client app is Outlook. It will not start popping system tray balloons on you until your Exchange has been unavailable for at least a few minutes.

So, the bottom line is - it is quite normal that some applications may become unavailable for a short period of time during VSS freeze process. This is just something you have to live with, no matter of backup solution you use (even legacy ones). Of course, you can always disable VSS processing and take the risk of backing up all your servers crash-consistent (approach that was not possible before virtualization). However, keep in mind that VSS is the only supported way to backup most of the Microsoft server applications anyway. So, you do want to create restore points with VSS enabled at least periodically.