none
General Newbie Questions for a Single Virtualization Server RRS feed

  • Question

  • I’m trying to setup DPM2010 on a stand-alone Virtualization Server. I expect we will do more of these with similar characteristics for additional customers.

    First, there are some excellent concepts in DPM such as syncs every 15 minutes, support for tape backup, support for Virtual Machines etc. There are a lot of new concepts and new terms here too, so forgive me if I sound naïve. I am finding as I experiment with DPM, and as I review some of the forum postings that DPM, is a different animal than classical backup technologies, and it also has a ways to go before it could be considered mature. This, plus a lack of complete, organized documentation, makes it hard to determine capabilities.

     

    This physical server has four W2K3 VMs which will be upgraded to W2K8 in the next year or two. One of the VMs is the domain controller and the host is a member server In order to run DPM. This is for a typical office environment consisting of SQL and non-SQL databases, Exchange, and Terminal Services.

    There is a primary disk containing the OS and DPM files, a V:\ Drive containing the virtual drives (which will probably be split in to smaller partitions to control disk positioning of the files, and a common software source partition that stores copies of all the CDs and other files needed to do a full restore of the physical server or VMs.

    The primary drive is about 836GB. The objective is to create a full backup on a LTO-5 tape (1.5/3TB). With one of these tapes, (or one plus incremental tapes) I would like to be able to do a full restore to a replacement server that is identically or similarly provisioned.

    We would also like to support user level recovery of files within their own VM environment.

    I am currently expanding the Storage Pool array from 1.9Tb to 3.7TB (issues are being dealt with in a separate post).

    The customer would also like to be able to create their own backup tapes in their remote office using an old server. Once we understand the bandwidth requirement and if we can get the bandwidth in there, we hope to be able to accomplish this by upgrading one of their old servers to handle DPM2DPM4DR.

    With that being said, I will probably want tape library backup at the datacenter anyhow, in case of latency problems in getting critical data to the remote server in a timely manner.

    Now the questions:

    Single Tape BMR Restore with Domain

    My understanding is that BMR / SS backups are actually handled by WSB on DPM’s behalf and that the replica is actually stored as a normal DPM replica. Further, that to access a BMR, it must first be restored to a partition, then actually restored to the C:\ drive using a Windows DVD and normal WSB restore.

    If this is the case, then on the new server, we would first have to install W2K8-R2 and DPM to restore the tape. (Alternately, we could keep a cheap 2.5” laptop SATA drive available to boot into on an emergency basis for the first part of the restore.) The tape would be restored in its entirety to a new Storage Pool array.

    In order to gain access to the backup, the temporary DPM server would have to belong to a domain and my understanding, is it would not have to belong to the original domain, or is that incorrect?

    Next, if the DPM server has to belong to the domain being restored, how do we resolve the chicken or the egg quandary, that at some point the DPM server will have to belong to the VM’s domain, but the domain is in a VHD in the DPM backup tape. Is there any way to have DPM support a single tape backup? If not, what is the bare minimum we would have to store OFF the tape? What would happen if we lost a valid copy of the DC altogether? Could the backups still all be fully restored? If we did keep a backup copy of the DC on an emergency restore disk, cold the tape be restored if we back-dated the computer clock, or would the domain have to “authenticate” the time before the restore was performed (Otherwise we would have to create a new DC copy every couple of months.)

    Interaction of WSB and DPM

    During setup and testing of this server, we have been doing WSB backups of all the setup work on the server and in the VMs onto a separate hard drive. Is this safe? Does DPM still keep its BMR data separately from the WSB partition? Is there any interaction between WSB and DPM that could mess each other up? Can we treat the DPM and WSB backups as two, completely separate and disconnected backups? Is there any reason we could not do WBAdmin or NTBackup backups at any time of the day or night while running DPM?

    Disk Space vs Tape Space

    Even though the DPM Storage Pool is huge, does the tape have to store all the data that is in the storage pool? For example, we only really need the data from the last sync or last Express Full. Can we turn off saving previous shadow copies to tape, but still keep them on disk?

    Virtual Machines

    We can add the V:\ partition (the one that contains all the virtual hard drives), AND we can add each Virtual Machine and all their drive letters to DPM for protection. Does DPM maintain two completely separate backup copies, or is DPM smart enough to store all the content just once? Does DPM do a WSB style backup for the file level backup and a DPM Agent style backup for the Virtual Machine type backup?

    Backing up Shares

    Is anything other than the details for recreating the share backed up? For example, on restore, is just the share itself restored or is any file content restored too?

    Storage Pool Allocation Sizes

    What is the best strip size to use for the Storage pool array? I understand DPM uses a 16K Cluster Size. Would a 16K strip size offer better performance or should it be larger like 64K or 1MB?

    DPM2DPM4DR

    When using a second DPM Server to protect the first DPM Server, can the Storage Pool on the second server be smaller, or is it the same size, or is it larger? Would it be possible to only maintain the last image from the last Recovery Point, or does it have to maintain all of the ones maintained on the primary DPM Server? Is there any problem with the first server doing some tape backups (either regularly or ad hoc) and the second server doing tape backups? (Customer wants at least month-end tapes at their office if that is all that the bandwidth can handle, but I’m more concerned with the ability to do a full tape restore if all the hard drives actually melted and the WAN latency is too great to maintain daily data at their site.)

    Thanks in advance for any answers provided or links to appropriate explanations.

    Bob.

    Wednesday, April 6, 2011 7:24 PM

Answers

  • Here goes (just keep in mind you should have a strategy for restoring each server, the DPM server and a disaster recovery scenario for if the whole site goes.

    Single Tape BMR Restore with Domain
    If you are not using tape to protect the domain controller for the domain then a secondary off-site DC and DPM server would be a good choice to get you back up and running. http://technet.microsoft.com/en-us/library/ff634186.aspx 
    If using tape then you basically would need to build a quick DC and DPM box to recover the BMR of at least one DC and then restore the DPM server into that domain and other BMR/data etc...... unless your tapes are certificate encrypted - but I'd suggest you don't do that for the scenario you've presented. You could then restore each servers information from tape as well as the DPM DB.

    Interaction of WSB and DPM
    Because DPM BMR uses WSB to create the BMR 'image' to stream to the DPM server the two are connected. What you need to be careful of is where a system state or BMR copy of the data is stored on the server (so DPM doesn't double up when it's protecting BMR). I don't know that WSB supports two concurrent backups of a server so you'd have to stagger backups so DPM and a WSB started backup don't conflict - but I don't see any real issues with using both methods to keep backups elsewhere.

    Disk Space vs Tape Space
    When you backup to tape using DPM it will basically do what every other tape solution out there does - create a copy of the data onto tape. This isn't the same size as the DPM storage pool (remember that DPM keeps both the replica and recovery point volumes on the storage pool = data size + change data blocks over retention period.
    So you're copy of data to tape will be the same as your data size

    Virtual Machines
    Just use the Hyper-V support for protection Hyper-V VM's available in DPM 2010 http://technet.microsoft.com/en-us/systemcenter/dm/ff632004

    Backing up Shares
    I believe 'Server Share' permissions and information is stored in the Windows registry on a file server (so protected by your BMR or System State). When you choose to protect data based on 'shares' in DPM you are protecting the data contained in that share, instead of choosing data on a DPM volume.

    Storage Pool Allocation Sizes
    This is more of a storage engineer question than a DPM question. Can't help you out on that one.

    DPM2DPM4DR
    See the link on DPM disaster protection from earlier. Basically the retention periods on the secondary can be different from the primary. In essence this means the DPM recovery point volumes can be smaller or larger, but the replica volumes will need to be able to store the same data that is being protected on the primary. Of course you don't NEED to protect all of the data sources that you are protecting on the primary out on the secondary but it's a good idea if you need that data and you are using this solution for DR.
    You can also modify the schedules for file data on the secondary, but for application data DPM relies on a 'Sync' schedule which cannot be greater than 24 hours. So depending on the schedule on the primary is how often the application data is sent to the secondary - but you can still reduce the recovery point retention if off-site storage comes at a premium to try to keep your storage usage down.

    Hope this is helpful.

    • Proposed as answer by MarcReynoldsModerator Friday, April 29, 2011 4:06 PM
    • Marked as answer by BobH2 Tuesday, October 4, 2011 5:14 PM
    Monday, April 11, 2011 7:35 AM

All replies

  • Here goes (just keep in mind you should have a strategy for restoring each server, the DPM server and a disaster recovery scenario for if the whole site goes.

    Single Tape BMR Restore with Domain
    If you are not using tape to protect the domain controller for the domain then a secondary off-site DC and DPM server would be a good choice to get you back up and running. http://technet.microsoft.com/en-us/library/ff634186.aspx 
    If using tape then you basically would need to build a quick DC and DPM box to recover the BMR of at least one DC and then restore the DPM server into that domain and other BMR/data etc...... unless your tapes are certificate encrypted - but I'd suggest you don't do that for the scenario you've presented. You could then restore each servers information from tape as well as the DPM DB.

    Interaction of WSB and DPM
    Because DPM BMR uses WSB to create the BMR 'image' to stream to the DPM server the two are connected. What you need to be careful of is where a system state or BMR copy of the data is stored on the server (so DPM doesn't double up when it's protecting BMR). I don't know that WSB supports two concurrent backups of a server so you'd have to stagger backups so DPM and a WSB started backup don't conflict - but I don't see any real issues with using both methods to keep backups elsewhere.

    Disk Space vs Tape Space
    When you backup to tape using DPM it will basically do what every other tape solution out there does - create a copy of the data onto tape. This isn't the same size as the DPM storage pool (remember that DPM keeps both the replica and recovery point volumes on the storage pool = data size + change data blocks over retention period.
    So you're copy of data to tape will be the same as your data size

    Virtual Machines
    Just use the Hyper-V support for protection Hyper-V VM's available in DPM 2010 http://technet.microsoft.com/en-us/systemcenter/dm/ff632004

    Backing up Shares
    I believe 'Server Share' permissions and information is stored in the Windows registry on a file server (so protected by your BMR or System State). When you choose to protect data based on 'shares' in DPM you are protecting the data contained in that share, instead of choosing data on a DPM volume.

    Storage Pool Allocation Sizes
    This is more of a storage engineer question than a DPM question. Can't help you out on that one.

    DPM2DPM4DR
    See the link on DPM disaster protection from earlier. Basically the retention periods on the secondary can be different from the primary. In essence this means the DPM recovery point volumes can be smaller or larger, but the replica volumes will need to be able to store the same data that is being protected on the primary. Of course you don't NEED to protect all of the data sources that you are protecting on the primary out on the secondary but it's a good idea if you need that data and you are using this solution for DR.
    You can also modify the schedules for file data on the secondary, but for application data DPM relies on a 'Sync' schedule which cannot be greater than 24 hours. So depending on the schedule on the primary is how often the application data is sent to the secondary - but you can still reduce the recovery point retention if off-site storage comes at a premium to try to keep your storage usage down.

    Hope this is helpful.

    • Proposed as answer by MarcReynoldsModerator Friday, April 29, 2011 4:06 PM
    • Marked as answer by BobH2 Tuesday, October 4, 2011 5:14 PM
    Monday, April 11, 2011 7:35 AM
  • I thought I'd add some answers after using DPM for awhile.

     

    Single Tape Restore
    A single tape restore is a long an elaborate process. First, install DPM using a temporary DC. DPM treats the tape as "foreign". Restore the real DC to a temporary location. Then reinstall DPM and join the real domain using the temporary copy of the DC. Then restore the host server and all the VMs. Then rejoin the host server to the real VM DC. This is a long process likely taking one to two days. It is much simpler if a backup domain controller is available on another machine. A physical backup DC can also provide a reliable time source. Its a pain, but because of the complexity, make sure you have performed a single tape restore at least once before going live so you can document all the issues and tools you will need if disaster ever strikes.

    DPM and WSB
    DPM needs to do a WSB backup within each VM and within the host IF a system state backup is run. A DC, for example, can live in a 25GB partition, but you will need to add 10GB if you do a DPM System State Backup. Both WSB and DPM can co-exist on the host and in a VM. 

    VM Disk Space
    A VM can be protected as a Child Partition Snapshot, baremetal or system state, as a remote machine or shared partitions and folders. The partition it is stored on can be protected too. Each instance maintains its own copy - so lots of disk space can be consumed in the DPM Storage Pool. Child Partition Snapshot is most flexible for a restore but only creates one Recovery Point per day.

    I have chosen to protect VMs as Child Partition Snapshots so I can restore a whole VM, individial VHDs, or just files or folders. I also do System State separately for the Primary DC. I protect the server DPM database and VM databases separately which provides better granularity for restores. Critical partitions and folders could be protected separately for granularity too.

    Backing Up Shares
    A Backup DC (or other machines) can backup to a share on the DPM Server using scheduled WBAdmin backups or WSB Backups. The backup folder is shared, and DPM can backup just the shared folder this way.
    Additional Tip
    Allocate way more Storage Pool Space than you think you will need. If you are building using high performance disk arrays, you can allocate one partition in the Storage Pool array for a host WSB backup which will be fast (or just use a standalone drive). Keeping a WSB backup for the host separate from the primary array provides for quick recovery of the primary array.
    When doing recovery work, having the additional disk space available on the Storage Pool array to restore a VM temporarily is really convenient. It means you don't have to shut down a production VM while you mess about with a recovered VM.
    (You can also temporarily convert a hot-spare drive to a restore drive if you are really stuck.)
    You can get by using (primary array size x4) but that's squeezing it. X6 or X8 is much better long term.
    Bob.

     


    • Edited by BobH2 Tuesday, October 4, 2011 5:22 PM
    Tuesday, October 4, 2011 5:14 PM