Just wanted to introduce myself. I’m a refugee from reddit currently as well, and strongly support not using them until they pull their heads out. Which probably won’t happen.
I digress.
I’ve been a ZFS user for … years. On Solaris, FreeBSD, Linux. I’ve used it to run a business and now my main media and NFS storage for homelab Promox cluster.
I also want to pose a question to the community, I’ve been doing some research in this area. I recently went looking for a place to back my data up offiste. I’m aware of rsync.net and other object storage solutions. Those all seem to be file-level backup and not zvol/vdev backups.
Is there any provider anyone’s aware of that offers the ability to do a full zvol/vdev? And the big question – if such a service existed, would you be interested in using it? If you’re willing to answer that question, and might be interested, what do you feel would be a reasonable cost for the service?
Glad to be here, and appreciate feedback on the question!
Rsync dot net offers direct zfs replication targets. It’s just a different offer from their primary, rsync/sftp/etc based offer. You can zfs send directly to it, use syncoid, whatever.
https://www.rsync.net/products/zfsintro.html
What you get is a full-on FreeBSD VM or jail (I don’t know which) with you as root. You can install software from repos, SSH in, etc. The only real limitation is an external firewall that blocks access to almost all inbound ports, but you can request manual exceptions if you need a port open that isn’t already.
I’ve been using the service for years now. It works very well, as long as you’re someone with the sysadmin chops to manage your own replication directly from the shell with no GUIs or hand-holding.
There was one that I found a while back. I thought it was rsync.net? But since you have to have access to the root level to push a zfs snapshot, the only way to do it was to basically give you a VM and some hardware. So they charged like $25 a month. I ended up just using file level backup since it was only a few dollars a month for what I need.
I always felt they should have written a send/recv preprocessor to isolate that more nicely, or just shim the send/recv code entirely into userland for this.
Thank you, I don’t know how I missed that. I need to do a better job of RTF website.
I can see that becoming a little bit more pricey, especially at a minimum of 4TB, and my datasets are … large, I have about 30TB of data I want to keep offsite.
When you think about encryption in particular, and not wanting people to see what’s in a dataset, the idea of replicating the whole zvol is pretty compelling.
I am using an S3 storage provider and backing up with Duplicati. Good client encryption, decent speeds.
The nice thing with this setup is that each bucket in s3 can have an arbitrary name, creating a user, group and policy per bucket with arbitrary names. API and access keys per bucket. The files in the bucket will get arbitrary names from Duplicati in 50MB chunks by default.
Even in the event someone gains access, its impossible to even recognize what is was let alone decrypt it.
