none
Huge Secondary Server Backups taking hours. RRS feed

  • Question

  • Hello.

    I have a setup with 2 (virtual) DPM Servers in separate Data Centres, connected by a fast WAN link. DPM is 2012 R2 v4.2.1417.0 (UR9). One DPM Server is configured as a Primary Server, the other is a secondary, backing up the data sources on the Primary Server. The data sources are a combination of domain controllers, SQL, Exchange and Web Servers, with a few Application Servers. They are spread across a number of separate untrusted domains, with no connectivity between the remote domains and none back to the Management domain hosting both the DPM Servers. All are VMs hosted on Hyper-V Servers (Win Server 2012 R2). No DPM Agents are deployed to any of the VMs, backup is solely at the VM level.

    Everything seems to be working on the Primary DPM server at the moment. But a number of the "replication" jobs to the Secondary server are taking a large amount of time and are transporting massive amounts of data across the WAN. This seems to be solely related to the Exchange and SQL Servers. For example, I have a backup of an Exchange server on the Primary that ran last night and copied approximately 28.5GB of data to its recovery volume. This ran at 18:00 last night and completed in 1:45 hours. I consider this to be reasonable, given the load on the Servers. I presume this 28GB is comprised mainly of exchange database, log and backup files.

    My problem is that the "replication" of that backup to the Secondary server does not copy just 28.5GB. The Secondary Server PG job is configured to run at 08:00 every 24 hours. This job has been running since 03:12 this morning and has copied nearly 600GB. So it looks like the Secondary job is replicating the entire VM, rather than changes to the VM (i.e. the 28.5 GB mentioned above). Checking the Hyper-V Host shows that this Exchange Server Storage has nearly 700GB used from an allocation of 2TB, spread across several partitions within the Exchange Server. I can't see why the replication job is running at 03:12 in the morning, not 08:00, I can only assume it is because replication jobs are taking so long and overrunning.

    My question is this. Given the way I am configured (VM Backups only) is this the reason why the whole VM is being replicated? This is inefficient for both Disk and Network traffic. Is there a way to amend what is being replicated so that it only transfers the smaller changes to the Secondary Server?  I have been using DPM for 3 years or so, and have never seen anything to indicate this can be done.

    I should also mention that I am starting to experience disk space issues on the Secondary DPM Server and may at some point need to provision more space, although this is going to be difficult for a host of reasons, so if I can reduce the amount of data transferred, this would be a huge benefit.

    Many Thanks to all

    regards

    John


    Regards, John

    Wednesday, August 10, 2016 2:39 PM