One of the last workloads to migrate for my current environment is a 3-node Windows Failover Cluster with File Server for general use roles. The 3 nodes are Windows Server 2012 R2 guests on an HPE ProLiant blade Hyper-V cluster connected to an HPE StoreServ 3PAR 7200 via iSCSI.
I needed a solution that would:
- Provide high availability and fault tolerance
- Support a minimum of two sites
- Integrate with Veeam and HPE StoreOnce with Catalyst deduplication
- Provide a path to cloud infrastructure and services
I’d been testing different solutions for about a year between balancing daily requests and team projects.
DFSR could work but can have sync issues and seeding takes some effort. And there must be something else out there. It would work with Veeam VM backups. Path to cloud was questionable.
Maybe Storage Replica. It provides high speed data replication and high availability if using a stretch cluster. Running a guest cluster in our Nutanix environment with shared storage has been problematic at best and frustrating. Veeam Agents can protect a failover cluster, but not Veeam VM backups. That means backups to a Veeam Backup Repository or SMB share and then backup copy jobs to StoreOnce. The Nutanix storage fabric capacity was sized for running workloads, so subscribing it for backup storage was not feasible and purchasing compute and storage from a campus vendor would be a yearly expense with limited RoI. HPE StoreOnce supports CIFS shares for SMB but performance testing displayed slow transfer rates for backup and backup copy. Too many integrations, too many opportunities for something to go wrong, and Storage Replica is overkill for user home directories and group folders. Path to cloud also not great.
It was a lot of time spent experimenting and waiting for Azure File Sync to go from Preview at Ignite 2017 to GA in July 2018. This looked like the good stuff. A file sync service that works with on-premises Windows file shares, moves the storage hub to Azure, uses Cloud Tiering to keep the on-prem VM disk footprint small. Bottomless file server? Yeepp…
I came home from Orlando with a mission:
>Deploy AFS in dev and learn how to monitor it.
>Estimate costs for LRS or GRS storage accounts for about 10 TB of files.
>Get it all backed up.
>Ship it.
And that was September and October, where I had a lofty goal of cutting over our file shares during fall break or winter recess. There were still governance and technical configuration issues to address. I did not see how Azure backups would meet our retention needs as it appeared to only support daily jobs and 60 days retention. That is unless you contact the Azure Backup Team and ask for their sample runbook to automate on-demand backups.
I needed a solution for storing a full copy of our data on-prem for governance and local backup. Support for deduplication and cloud tiering came later with the February 2019 v5 Agent release. Prior to that I was concerned with the capacity needed on our two Nutanix Hyper-V clusters. I could have a full copy of a server endpoint on one cluster, a tiered endpoint on another cluster. More complexity.
This is a compromise but at least for proof of concept with some production lifetime, I like it. I combined 2.5" 5 TB Seagate BarraCuda spinning disk with HPE mixed use enterprise SSD in a ProLiant DL380 G9 to create a Windows Storage Pool with Tiered Storage. The performance is fair considering the cheap consumer HDD. The Virtual Disk and its volume hold the VHDX files for a VM whose sole purpose is to sync Azure Files. This is not a user facing server and there are no shared folders. The Windows Server 2016 host and Veeam 9.5u3 VM backups means Resilient Change Tracking makes quick work of protecting data.
More work to come with data migration, analytics, go live and troubleshooting.
jcengelking.dev 