Friday, October 27, 2017

Protecting Hyper-V VMs

This section covers what can be protected, different deployment topologies, and performance and scale numbers. Hot backups of Microsoft workloads (for example, SQL Server, Exchange Server, SharePoint Server, and file servers) and Linux workloads is supported. Note that you get app consistent backups for the VMs running Microsoft workloads while you get file-consistent snapshots for the VMs running Linux workloads due to no Volume Shadow Copy Service (VSS) support in the guest.

DPM can be used to protect VMs hosted on the following:
 Standalone Hyper-V hosts that use local or direct-attached storage. Note that this option does not provide continuous availability and is not recommended for production deployments.

 Hyper-V cluster with storage on Server Message Block (SMB) shares backed by an SOFS cluster. This deployment type is referred to as “Hyper-V over SOFS.”

 Hyper-V cluster with the virtual hard disks stored on Clustered Shared Volumes (CSV). This deployment type is referred to as “Hyper-V over CSV.”

Protecting Hyper-V over SOFS
Hyper-V over SOFS configuration (with Storage Spaces) enables cost-effective, highly available, scalable, and flexible storage solutions for business-critical virtual deployments. It leverages industry-standard storage and allows customers to use Windows Server for highly available storage that can cost-effectively grow with demand. In this configuration, compute and storage are decoupled, and you can independently scale one without the need to scale the other. This configuration provides the lowest acquisition and operations cost. It allows highly available VMs, continuously available file servers, and fault-tolerant storage. The only challenge with this configuration is hardware setup and software installation unless you decide to go with the Cloud Platform System option.

Here are some considerations when using Hyper-V over SOFS and what they mean from a backup perspective:

 The file server must have Windows Server 2012 R2 with the new SMB 3.0 protocol. In order to get scalable backups with DPM, you cannot use non-Microsoft file servers that implement the SMB 3.0 protocol (because the DPM agent doesn’t work with these file servers).

 You must have separate failover clusters for Hyper-V and for the file server. DPM agents should be installed on the Hyper-V cluster nodes and on all the storage nodes (since the storage server is also clustered). You’ll need full-share and folder-level permissions for the local $ account of the file server on the SMB share.

Now consider the failure scenarios. What if the compute node goes down? There is no impact to the protected VMs on that node since the tracking is on the storage node. If there is a DPM VM on that node, all running jobs fail and are re-tried automatically.

What if the storage node goes down? All VMs that are touched by this storage node (since the last backup) go into CC mode since the tracking is on the storage node. In Cloud Platform System, where one rack can host up to 2,000 VMs on a four-node SOFS cluster, roughly 500 VMs go into CC mode when one storage node goes down.

How do the scale numbers look? In a scale test conducted by the DPM Product Group, continuous daily backups for three weeks (using virtualized DPM servers) were taken where the workload running inside each of the VMs was spread across multiple I/O profiles (SQL OLTP, Exchange, File Server, Video Streaming, and SQL Decision Support System). The guest operating system used for the protected VMs was Windows Server 2012 R2.

Scale testing with each DPM server protecting between 50 to 250 VMs was performed. DPM VMs were deployed in scale-out configuration to protect VMs from the same Hyper-V cluster nodes. Results were pivoted around the following criteria:

 Backup success rate per day This signifies the percentage of VMs having successful backups in a single day.

 Overall backup success rate This signifies overall percentage of successful backups across all VMs for a three-week duration.

More than 98 percent success was achieved for both the metrics. It also implies that there were more than 20,000 jobs that ran successfully during this three-week duration. The few errors encountered were due to known auto-recoverable failures, such as “Out of storage space” and “Retry-able VSS errors.”

Protecting Hyper-V over CSV
Hyper-V over CSV (backed by SAN) is the most predominant deployment where VMs are hosted on a Hyper-V cluster with CSV storage. There is no limit to the number of disks a Hyper-V cluster can be configured to use, which allows much flexibility in designing the storage architecture of Hyper-V host clusters. For backing up the VMs, the DPM agent is installed on each cluster node, and you get reliable backups at scale with the latest version of DPM 2012 R2. There are no more host-level volume snapshots; it calls guest-level VSS to get application-consistent backups. DPM supports express full backups and parallel backups.

Now, consider the failure scenarios. What if the compute node goes down? All protected VMs on that host go into CC mode since the filter tracking is on that node. If there is a DPM VM on that host, all running jobs fail and are re-tried automatically. There is no impact on the jobs that have succeeded, and the jobs in queue continue to be in the queue even after the DPM VM moves to the new node in the cluster.

Source of Information : Microsoft System Center

No comments: