File-Level Backups

[peteWine]

I have a feeling this has been discussed before, but I was unable to find anything using the search function. Also, I am not an IT professional, so bear with me!

Sort of a unique situation here. We have a Server 2003 Terminal Server (30+/- users) VM running our bookkeeping/accounting/invoicing software and, prior to virtualization, encountered regular corruption on the database files used by the program. So far, we have not encountered any corruption since virtualization, but it has only been a few weeks. There are a few dozen of these files in MDF format, and I would like to implement rolling backups of these key files - one 10 minutes old, one a day old, one a week old, to simplify and speed DR in the event of corruption.

The issue I'm having with backups is that the DB MDF files are locked for the time they are being written to (hundreth of a second) or read (report being generated, may be several minutes). I need an unobtrusive backup method that will not cause corruption itself!

Would snapshots be an option in this case? I see no way of implementing them on a folder level, so I am assuming the answer is no. Input anyone could provide would be greatly appreciated.

Thanks!

Pete

[Lakshmi_Gayatri]

Backups do not corrupt data. Snapshots can be used at a VM level. You can use VMware Consolidate Backup etp take fil backups of your VMs or deploy any backup software which has VMware integration so as to help you take file backups.

[idle-jam]

backup that utilize VSS (it should work with most backup software, make sure your vss is runnign fine) snapshot for your MDB will work great.

[AttilaSzamos]

Try Veeam - Backup and Replication v5, it is for backing up vSphere VM-s. It uses VSS for backup and for restore operations as well.

--

Attila

http://blog.szamosattila.hu

[noahzenzen]

You might want to try looking at www.inmage.com

[peteWine]

Thanks for the replies, everyone. I do currently use Veeam to replicate and backup our VMs, but was hoping for something lighter. The issue with corrupt is not a matter that the backups themselves create corruption, it is a matter of the database files being locked during writes which may be causing corruption as multiple users try to hit it.

I do have a testing environment, so I'll give some of these suggestions a whirl! I'll let everyone know what I decide.

[Lakshmi_Gayatri]

Simpler file mount can be achieved with VCB as well but in VCB you cannot just choose a VMDK but all the VMDKS get mounted on to the VCB proxy.

If the database you are using is of Microsoft, then you can use the VSS feature while snapshotting.

Regards

LG

[ThomasMc]

Are you using a backend/frontend setup?

Any chance you could upgrade the db to SQL?

SQL would save you from a world of hurt

[squebel]

You could always take a non-quiesced snapshot of the vm which will not invoke VSS commands to freeze the filesystem and the databases. You could then backup the virtual machine while the snapshot is running. Obviously your OS and db's will be crash-consistent, but if you are doing local backups of the database only with transaction log backups, you should be fine. You could then restore the entire virtual machine and then use your local backups to recover the database.

[admin]

It really doesn't matter what backup solution you use here... all of the backup solutions that integrate into VMware either use vcbmounter which has been replaced by the vStorage API. All integrated backup products do the same thing, tell the ESX server to perform a snapshot on the VM which then free's up the parent virtual disk to be backed up.

When the snapshot occurs there are three methods to try and get consistency at the point of snapshot. The first is to try and invoke Microsoft VSS to halt writing to the disks so that we don't get half written data at the point of the snapshot. If this fails the next step is to try the Sync driver which is also installed as part of VMware Tools. The Sync driver will attempt to force Windows to flush all unwritten data from memory and actually write it to disk. Once this happens the snapshot can then occur. If both of those fail the ESX server looks to run the freeze batch file that is located in the Windows folder on the hard disk. If you have edited that to stop any services, it will allow services to stop before the snapshot is taken and then have the services restart with the post freeze batch file.

There's not going to be one solution that is better than any other as the most preferred solution, using VSS, is something that is made or broken inside the guest itself and has nothing to do with the backup solution calling for the snapshot. Make sure Microsoft VSS is happy and your backups with run fast and consistently. Make sure you have Change Block Tracking working and things will go even faster for you.

cd.