VMware backups need to be planned carefully while taking into consideration your environment design. You might be backing up data for various reasons, with each backup requiring a different strategy, such as recovery of lost data or archiving. This chapter provides help with the implementation of snapshots, which can be very useful in data recovery policies. It includes the following topics:

A snapshot is a point-in-time copy from data, which allows the administrators to recover the data in that specific point.

That technique is useful with virtual machines, because it provides the ability to recover a server to a specific point whenever needed. If a risky change is going to take place on a certain server, a snapshot can be taken just before the change begins. If anything goes wrong during the implementation, it is not necessary to follow the traditional restore approach of servers. That is, you do not need to install the operating system from scratch, install the backup, restore the software, and restore the data. The only step needed is to restore to a previous point in time. The server is up and running again in a matter of seconds or minutes, depending on the amount of data to be reverted.

Formerly, it was considered a best practice to separate the real server‘s data from transient data, such as temporary files and swapping partitions. But as virtualization implementations became more mature, this practice changed. Data separation does not add enough benefits to justify its implementation, because it changes the way the servers are configured.

The major benefit of data separation is reducing the amount of data to be stored on snapshots and replicated to remote locations in case of disaster recovery (DR) implementations.

However, keeping all the pagefiles in a single location creates a single point of failure, because if it fails, all the virtual machines are affected. The separation also adds an administrative burden. It requires the reconfiguration of all servers to point to a new disk in a new “transient datastore,” responsible to hold that temporary data.

For all these reasons, the new best practice is to keep the transient data stored with the server‘s data, providing a centralized management of the entire solution.

VMware snapshots are a valuable tool to manage the environment, but they do not cover all the restore and performance possibilities.

To understand this idea, it is necessary to understand how snapshots work in the VMware world. Basically, the VMware snapshot system locks the virtual disks (.vmdk) at the moment of the snapshot. All new information from that point in time is not written to the .vmdk, but to a file created on the same directory as the .vmdk. If the virtual disk is named C.vmdk, a file named C-000001.vmdk is created, and all new information is written on it instead of the C.vmdk. For each read or write operation, it is necessary check two different files, the original and the -00000x.vmdk, to complete the operation. It might cause serious performance delays, especially on high disk I/O virtual machines.

Because all the new information is never committed into the .vmdk, the snapshot file grows indefinitely. It can take all the available space on the datastore where it resides, which can cause a crash of all VMs that share the same datastore.

Another reason to avoid maintaining long term snapshots and keep taking new ones is the access of multiple files to get the information need. If only one of those files gets corrupted for any reason, you lose all the information stored on that .vmdk. You must then consider how you can restore the data.

One last point to consider is that snapshots do not cover complete disaster recovery. If both the primary and secondary sites have serious problems as a natural catastrophe, there would be no data to restore from, as opposed to the common approach of keeping backups stored on media on a remote site.

Keeping VMware snapshots of virtual machines is a great point-in-time recovery option, but might cause performance issues. Next we consider how to take advantage of these features while maintaining a system running without impacts.

The solution is to integrate VMware and storage snapshots. In that way, a VMware snapshot is taken, followed by a storage snapshot from the volume containing that VM, and then the VMware snapshot can be deleted, avoiding performance impacts.

The following sections show you how to implement that solution.

To use snapshots as a solution, take a virtual machine snapshot, then take a snapshot from the volume where the LUN and its respective datastore. Then remove the virtual machine‘s snapshot.

Use vCenter to take snapshots, as shown in the following steps:

After the VMware snapshot is completed, take the N series snapshot. To take a snapshot of a volume where the LUN of a datastore reside, use the following steps:

Follow these steps:

In a production environment, you can automate the snapshot process. vCenter can be used to schedule snapshots as follows:

The N series component can be scripted as indicated previously by using the command”

To perform that action using the GUI, open IBM N series System Manager, complete the following steps:

Open IBM N series System Manager, select the Storage → Volumes, then select the volume, click Snapshot → Configure. It will display a number of options as shown in Figure 9-11. It is important to set the number of snapshots to keep in accordance with your storage capacity. Also, schedule the snapshot to occur when the production utilization is low to avoid bottlenecks. Click Apply.