Continuous Backup for Exchange Server 2010 with DPM 2012 (Part 3)
If you would ike to read the first part in this article series please go to:
Other Exchange 2010 Roles
So far we’ve created protection for the Exchange mailbox databases, but what about the other Exchange roles?
In order to fully protect an Exchange Server 2010 infrastructure, other Exchange Server roles such as the Hub Transport and Client Access Server should also have a recovery plan, in the event of failure. The configuration data for these roles is stored within Active Directory, but the working data is stored in the file system. To make a list of what to backup in these roles, I recommend reading the following articles:
If DPM 2012 is used for the protection of servers hosting the Hub Transport and Client Access server roles it is possible to perform a complete bare-metal recovery of the servers hosting these roles. This recovery can be performed either to physical hardware or a virtualized infrastructure. It will also take significantly less time than reinstalling and configuring the host operating system on a replacement server in preparation for running the SETUP.EXE /m:RecoverServer Installation recovery routine.
After knowing what to backup in each role, configuring a Protection Group for other Exchange roles should be pretty straightforward. Figure 1, Figure 2 and Figure 3 depict an example, showing parts of the Protection Group wizard. The main difference in this case is in the Select Group Members page: instead of selecting mailbox databases, one must select the appropriate file systems, System State and Bare Metal Recovery.
1: Select Group Members
2: Select Data Protection Method
3: All Protection Groups
Performance should be a top priority for any Systems Administrators, especially when it concerns protecting data and computers. There are several factors that can impact performance, such as the hardware specs of the DPM server, network speed, hardware specs of the protected computer, size of the protected data and the rate at which the protected data changes.
Let’s take a deeper look into how all these factors influence the final performance and what measures can be taken to mitigate potential risks.
DPM Server CPU and Memory
On the DPM server, three processes can impact performance:
- DPM protection agent (MsDpmProtectionAgent.exe): DPM jobs affect both memory and CPU usage by the DPM protection agent. It is normal for CPU usage by MsDpmProtectionAgent.exe to increase during consistency checks.
- DPM service (MsDpm.exe): the DPM service affects both memory and CPU usage.
- DPM Administrator Console (an instance of Mmc.exe): DPM Administrator Console can be a significant factor in high memory usage. You should close it when it is not in use
Under normal conditions, memory usage for the DPM instance of the SQL Server service (Microsoft$DPM$Acct.exe) is expected to be comparatively high. This does not indicate a problem, as the service normally uses a large amount of memory for caching, but it releases memory when available memory is low.
DPM Server Storage I/O
Since the storage subsystem can constitute a bottleneck, the following 2 measures can improve performance:
Adding disks to the storage pool and reallocating the replicas across the storage pool can help reduce disk queue length.
Using striped volumes can increase disk throughput to deal with disk bottlenecks.
DPM Data Transfer Operations
DPM data transfer operations have a huge impact on system and network resources. Let’s look at each of these operations individually and see what can be done to mitigate the risk of a bottleneck:
- Replica creation: Typically, the performance of the replica creation will be limited by the speed of the network connection between the DPM server and the protected computers. That is, the amount of time that it takes to transfer a 1-gigabyte (GB) volume from a protected computer to the DPM server will be determined by the amount of data per second that the network can transmit. The impact of replica creation on network performance can be reduced by using network bandwidth usage throttling.
- Synchronization: Synchronization is the process by which data changes are transferred from the protected computer to the DPM server and then applied to the replica of the protected data. For protected Microsoft Exchange data, synchronization transfers an incremental Volume Shadow Copy Service (VSS) snapshot.
Each synchronization job consumes network resources and can therefore affect network performance. The impact of synchronization on network performance can be reduced by using Network Bandwidth Usage Throttling and On-the-Wire Compression.
- Consistency check: A consistency check is the process by which DPM checks for and corrects inconsistencies between a protected data source and its replica. The performance of the protected computer, DPM server, and network will be affected while a consistency check is running, but the impact is significantly lower than initial replica creation, because only the changes and checksums are transferred. Consistency checks should be performed during off-peak hours.
- Express full backups: The impact of an express full backup operation on performance and time is expected to be less than the impact of a full backup because DPM transfers only the blocks changed since the last express full backup.
Network Bandwidth Usage Throttling
Network bandwidth usage throttling causes jobs to use less bandwidth, but they take longer to complete. To enable it, follow these guidelines:
In DPM Administrator Console, click Management on the navigation bar. Click the Agents tab.
In the Display pane, select a server. In the Actions pane, click Throttle computer.
Click Enable network bandwidth usage. You can configure network bandwidth usage throttling separately for work hours and non-work hours, and you can define the work hours for the protected computer. Work hours and non-work hours use the time zone of the protected computer (Figure 4).
On-the-wire compression decreases the size of data being transferred during replica creation and synchronization but increases CPU utilization on the DPM server and the protected computers. The amount of compression and improvement on network performance depends on workload.
Compression is enabled for a protected computer and applies to replica creation, synchronization, and consistency check operations. Recovery jobs also use compression.
To enable on-the-wire compression:
- In DPM Administrator Console, click Protection on the navigation bar. In the Actions pane, click Optimize performance.
- On the Network tab, select Enable on-the-wire compression. To apply your changes, click OK (Figure 5).
5: On-the-wire compression
DPM can be centrally monitored using System Center Operations Manager (SCOM). You use the Alert Publishing option only if you have chosen to centrally monitor your DPM servers in SCOM. This option synchronizes the DPM alerts that are displayed in the DPM Administrator Console with the Operations Manager console.
To publish existing DPM alerts, in DPM Administrator Console, click Options. In the Options dialog box, on the Alert Publishing tab (Figure 6), click Publish Active Alerts, and then click OK. On the pop-up message window (Figure 7), click OK.
6: Alert Publishing
7: Publish Active Alerts
After enabling the Alert Publishing option, all existing DPM alerts that might require a user action are published to the DPM Alerts event log (Figure 8). The Operations Manager agent that is installed on the DPM server then publishes these alerts to SCOM and continues to update the console as new alerts are generated.
8: DPM Alerts events
You can also configure DPM 2012 to send notifications by e-mail of critical, warning, or informational alerts, and the status of instantiated recoveries. If you plan to enable this feature, you must first configure the SMTP Server.
Follow these procedures to subscribe to notifications:
- In DPM Administrator Console, on the Action menu, click Options.
- In the Options dialog box, on the Notifications tab, do the following:
a. Select the types of alerts about which you want recipients to be notified.
b. Under Recipients, type the e-mail address for each recipient to whom you want DPM to send copies of the notifications (Figure 9).
9: Configuring Notifications
- Go to the SMTP Server tab (Figure 10), type the SMTP server name, the SMTP server port, and the e-mail address you want to display. In the Authenticated SMTP server area, type a user name and password in the appropriate boxes.
10: Configuring SMTP Server
- To test the SMTP server settings, click Send Test E-mail, type the e-mail address where you want DPM to send the test message, and then click OK.
Backstage of Backups
Once the initial replication to the DPM server of the Exchange databases is made, synchronization will occur with the frequency that was configured for the short-term recovery points. The protection status of each storage group will be shown in the DPM Administrator Console (Figure 11).
11: Synchronizing Exchange Protection Group
The Exchange servers we’re using have a dedicated mailbox database each, and also a copy of the Contoso-Users-01 database, which is the mailbox database being protected by the DAG (Figure 12).
12: Storage Groups and Mailbox Databases
Since this is not a production environment, all the Exchange data files (log files and databases) are located on the C: drive (Figure 13). Notice that there are 293 files in the folder.
When an Express Full Backup occurs, all the log files are pruned after a successful synchronization, since this is actually a full online backup (Figure 14).
13: Exchange data files
14: Exchange data files after Express Full Backup
If you remember part 1 of this series, it has been said that the synchronization process uses the Exchange VSS Writer. If we take a closer look to the Application Event Log of the protected Exchange server, we’ll notice a couple of events related with the VSS Backup (Figure 15, Figure 16, Figure 17 and Figure 18).
15: VSS Writer event 2021
16: VSS Writer event 2027
17: ESE Backup Event 961
18: VSS Writer Event 2046
If you open Server Manager, you can see the volumes created by the Protection Group (Figure 19), which are configured as mount points. If you right click one of them and select Change Drive Letter and Paths you can actually see the full path where those volumes are mounted to (Figure 20). This same information is also displayed in DPM Administrator Console, in the Protection task area, select a mailbox database and click the Replica Path hyperlink on the bottom pane (Figure 21).
19: DPM Disk Allocation
20: Volume Drive Letter and Paths
21: Details of Replica Path
Using Windows Explorer, you can browse to the location of those replicas (Figure 22), although I would not recommend that you mess with the files on those folders.
You will notice that the installation path of DPM contains three folders in the Volumes directory:
<Install Drive>\Program Files\Microsoft System Center 2012\DPM\DPM\Volumes\DiffArea - contains mounted shadow copy volumes that store the recovery points for a data source.
<Install Drive>\Program Files\Microsoft System Center 2012\DPM\DPM\Volumes\Replica - contains mounted replica volumes.
<Install Drive>\Program Files\Microsoft System Center 2012\DPM\DPM\Volumes\ShadowCopy - contains local backup copies of the DPM database.
22: Browsing the Volume Mounting Point
When using DPM protection of your Exchange servers, you should be aware of the following additional considerations:
- Adding mailbox databases to the server. If you create or add new mailbox databases to a protected storage group on an Exchange Server computer, these databases will be automatically added to the DPM replication and protection. Incremental backups will fail until a full backup is completed
- Adding storage groups to the server. If you create or add new storage groups on a protected Exchange Server computer, these storage groups will not be automatically added to a DPM protection group. You must modify the Protection Group to include the new storage groups.
- Changing storage group or mailbox database file paths. If a protected database or log files are moved to a volume that contains data that is protected by DPM, protection continues. If a protected database or log files are moved to a volume that is not protected by DPM, an alert is displayed and protection jobs will fail. To resolve the alert, in the alert details, click the Modify protection job link and then run a consistency check.
- Dismounting mailbox databases. If you dismount a protected mailbox database, that protection job for that particular database will fail. The replica will be marked inconsistent when the next express full backup is run by DPM.
- Renaming storage group and mailbox databases. If you need to change the name of the storage group, follow this procedure to ensure continued protection: a) stop protection with retain data and b) reprotect the database. Until you reprotect the database the backups will continue to work but mailbox enumeration will fail.
- Renaming an Exchange server: DPM uses the computer name as a unique identifier for replicas, recovery points, DPM database entries, reporting database entries, and so on. You cannot do change the name of a protected computer and continue protection without disruption
- Moving databases between storage groups. If you move a mailbox database from one protected storage group to another, it will continue to be protected. If you move a mailbox database from a protected storage group to a storage group that is not protected, the mailbox database will no longer be protected once a consistency check has been performed. If you move a database from a storage group that is not protected to a protected storage group, it will automatically become protected once a consistency check has been performed. In all these cases, you must run a consistency check on all protected storage groups involved in the move once the mailbox database moves are successfully finished.
After spending 3 parts explaining the setup, configuration and operation of DPM 2012, the next and final part of this article will be dedicated to the recovery process, the most important feature of any backup solution.
If you would ike to read the first part in this article series please go to: