AWS - Any way to have cross-AZ redundancy with EBS

Question

I'm looking for a redundancy solution for EBS. Ideally we would have an EBS volume in each AZ - in something similar to a master-slave configuration. Upon the failure of the primary EBS volume, all reads/writes would be directed to the secondary volume.

The volume is to store user data for the application, no OS or application files or anything like that.

The EC2 instances that will mount this volume are running centos, and I haven't looked into any sort of linux solution just quite yet. If there isn't any amazon solution, perhaps there is a linux one that you know of?

It just doesn't make sense to me to have a multi-az deployment if we are stuck with a critical EBS volume in a single AZ.

Thanks!!

Gaming

Answer 1

No. Amazon EBS volumes reside only in one AZ. Snapshots are stored across AZs, but not 'live' volumes.

Your use-case sounds suitable for Amazon Elastic File System (EFS) for Linux or Amazon FSx for Windows File Server. They are shared network volumes that are replicated across AZs and can attach to multiple Amazon EC2 instances simultaneously.

I like to think of EBS like a directly-attached drive and EFS/FSx as a network drive. (Those aren't accurate descriptions, but they fit my mental model.)

Nasho

Definitely sounds like a use case for EFS, depending of course on what exactly "user data" means. – Reims

Answer 2

As John pointed out, you could use EFS or FSx for Windows. Another option would be to use FSx for NetApp ONTAP (FSxN). It is a managed service like those two, however, it is not as expensive and is much more performant, providing latencies in the sub-millisecond range. It also supports a Multi AZ deployment, with automatically transparent failover. Since you are already using EBS, you could probably easily translate that to iSCSI LUNs. Or take an easier route, and simply create NFS or CIFS volumes.

Maintop