Can google cloud replicate filestore disks to read-only disks in another region like AWS EFS can?

Question

We have a client that insists on using Google Cloud. They also want their application to be synced across regions.

In AWS, you can have an EFS (a disk drive, basically) automatically sync to another read-only EFS in another region. From this document.

Amazon EFS creates the destination file system with read-only permissions... the replication process continually keeps the destination file system in sync with the source.

I'm wondering what analogous service Google cloud has to replicate disks to a read-only disk in another region? The read-only disk would also need to be able to be mounted on any VM instance that started up (at the /var/www/html mount point).

Since this is nowhere in the docs, what is the best setup to build functionality for continuous replication of filestore across regions?

Answer 1

AWS EFS doesn't present a disk, but presents a shared filesystem over NFS. This distinction is important because while replicating a disk between two regions has benefits, if the filesystem doesn't support shared access (and I don't know one that does) it can't be mounted read-only in the remote region.

On Google Cloud the comparable service available today that supports cross-site regional replication with NFS protocol access is Google Cloud NetApp Volumes. With this Google service you would create a volume in your primary region, and then create a volume replication with your remote region. The volume can be mounted over NFS read-write in your primary region, and read-only in your remote region.

If your replication requirements are less frequent, and you have a single VM in each region, you could also use a local disk in each region and replicate at the file level. The Storage Transfer Service and a File system source and sink could be used, or rsync if you want a self-managed solution.

And a last idea for you might be GCSfuse which allows you to mount a Google Cloud Storage bucket as a filesystem. If you have a webserver with small assets it's not a good fit, but for workloads that are reading multi MB size files or saving out the same it could work.