Although application consistency is more often discussed in the context of virtual machine (VM) backups, it’s just as important to consider a strategy for fail over replicas. In fact, if you are both backing up and replicating the same VM(s) it becomes even more critical. With Veeam, the configuration for application consistency (and the GUI windows for doing so) are the same for replication as for backup. Therefore, much of the info here will be in reference to backups even though this is a post in a series about replication only.
Overall, this post touches on some of the options and strategies administrators have when planning to provide application consistency for both Windows and Linux VMs. Various links to KB articles, blog posts, and white papers are provided for additional reading.
This post is part of the VMETC.com From the Field Series on Veeam v6 Replication
The rest of this post is divided into 4 sections:
- Agentless options for Windows VMs
- Agentless options for Linux VMs
- VM Cluster nodes
- Should you truncate logs for a replication job?
A good download for starters
A few years ago Veeam presented Considerations for backing up Exchange and Oracle for a VMware User Group. These are popular applications, and I refer to this information in the rest of this post. The methods discussed for each provide a great example of the process and strategy involved in both Windows and Linux VMs. Download a copy here:
Veeam Backup and Replication Considerations for Exchange and Oracle (7431)
Agentless options for Windows VMs
Microsoft’s VSS framework makes quiescing applications running in Windows VMs a relatively easy process for the replication software without the need for installing agents in the VM. In short, code already exists to make the application consistent and put the OS in a recovery mode necessary to safely, and properly, restart the VM. For applications like SQL, Exchange, Oracle (when running on Windows), and Sharepoint, the VSS framework can even be leveraged to commit transaction logs. The VSS framework provides 2/3 of the solution. Veeam Backup and Replication provides the remaining pieces to ensure reliable application consistency.
As shown in the screen shots below, enabling transaction consistent replication with Veeam is as easy as choosing a checkbox and providing VM local admin credentials.
Veeam Backup & Replication – Step 12. Enable Application-Aware Image Processing
Some more recommended reading if you need to understand more about VSS:
What is Windows VSS & why you should care | Mr. Backup Blog
VMware and Microsoft VSS: What You Need to Know
Agentless options for Linux VMs
Applications running in Linux VMs require a little more set up today outside of the Veeam GUI, but the minor effort is well worth the time. The answer is to utilize VMtools and their ability to call pre and post thaw scripts.
Here are a few well documented how to examples from other IT Pros for the most popular Linux apps – MySQL and Oracle:
How to Quiesce Linux Virtual Machines with Veeam Backup – MySQL | HiperLogic
On Windows, Veeam can do application consistent backups through the Veeam Microsoft Volume Shadow Copy Services (VSS) integration for VSS aware applications. Unfortunately Linux does not have a system like Microsoft Volume Shadow Copy Services (VSS) that allows applications to be made application consistent by backup applications like VMware vDR or Veeam Backup.A simple workaround is to leverage VMware tools ability to call custom freeze/thaw scripts during a quiesced snapshot to enable customers to do the required application quiescence.The example [in the post] is for MySQL, for other applications you need to know the application specific command to quiesce the application and modify the scripts below appropriately.
Follow the link above to see the specific scripts for the MySQL example
Hot Backup of MySQL on a Linux VM
This whitepaper written by Veeam also shows a detailed process for MySQL VMs and provides sample scripts
Moving on, there are 3 common strategies for Oracle as outlined in the PDF copy of the VMUG presentation (downloadable above) :
The PDF provides the technical details of all 3 methods for Oracle running in Linux.
A possible solution of the future for Linux VMs may be the experimental vmsync option of the VMware tools. If you pay attention when installing VMTools for Linux you will notice the option for vmsync with a default answer to leave it disabled. vmsync seems to be the “VSS-ish” solution to enable the future GUI config of non-windows application quiescing.
Although I have not personally tested it, others have blogged about their experience enabling vmsync.
Filesystem consistent Linux Backups with VMware | Virten.net
Solution 2: vmsyncAs mentioned above VMware has created a sync driver that allows to create consistent backups. Unfortunately i couldn’t find any information about this driver and it is disabled by default, giving a small explanation:[EXPERIMENTAL] The VMware FileSystem Sync Driver (vmsync) is a new feature that creates backups of virtual machines. Please refer to the VMware Knowledge Base for more details on this capability. Do you wish to enable this feature?This comment emphasizes that this feature is not supported at the moment. I also couldn’t find this Knowledge Base details.
Replication of VM Cluster nodes
Unfortunately, agentless, host based replication solutions like Veeam cannot integrate with a clustered application. What you are replicating is the entire VM for one or more nodes. This means that the quorum and shared configuration of the cluster must also be replicated to the fail over location. This is often done outside of Veeam unless the quorum is somehow configured within a virtual disk that can be replicated. This is rarely the case. If you are able to account for all of the shared storage and networking at a DR site, then failing over may work.
I would suggest first analyzing how your cluster behaves when a node fails. What are the necessary pieces in your virtual infrastructure to keep the application available to the users running with a failed node? Essentially, the application cluster fail over state is what you need to replicate. You may not need to replicate all nodes, for example. If you only need a single node for the application to operate then save the storage space and the bandwidth and only replicate one VM node. Better yet, replicate the passive node if possible to lessen the impact on production while the job is running.
Should you truncate logs when replicating?
The gotcha with replicating VMs that you are already backing up could be that you introduce a double log truncation schedule from 2 different jobs. Veeam replication jobs are independent of backup jobs and only provide cross intelligence for schedule queuing and conflicts.
To avoid the double schedule simply disable log truncation for the replica jobs (select “do not truncate logs” in the Veeam screen shot from earlier). Since Veeam also provides file level restores from replication jobs, you might consider only replicating VMs and skipping the backups for certain VMs.
Summary
As always with virtualization, the total solution for application consistency requires the cooperation of multiple technologies. A host based replication product such as Veeam will act as the “conductor” of the over all process, but one or more of the “instruments” in the DR “concert” may need to be fine tuned. This post hopefully provided several examples to help you get started.
Posted with Blogsy
