BackBlaze B2 Update Question(s)

[KazuyaDarklight]

v11.2 - U3 on Primary, U4 on Backup.

We have 2 units. "Primary" and "Backup", the Primary runs normally and Backup takes in snapshots from Primary via Replication Tasks. We want to use B2 for our offsite backup/DR. With that in mind, am I best served by running the Cloud Sync on Primary, pointed at the actual raw data (maybe with retention turned on in B2), or Backup with the Sync just scooping up all the snapshots?

Any recovery is going to be held up by our ISP speed but while I like the idea of Backup handling all backup duties in principle, the Snapshot scenario seems like it would be trouble since we would need to download ALL the snapshots in order to start working with any of our data. Correct? We wouldn't really have the chance to prioritize particular directories?

If that is so, then my follow up question would be, if I point the sync task at the raw data and turn on version retention in B2, when I need to pull data back later, is it going to automatically pull the latest version of the files or is it going to download all the different retained file versions?

Thanks for your time.

 

[dlavigne]

Did you decide upon a solution for this?

 

[KazuyaDarklight]

Due to my current ignorance, combined with pressure to get things moving, I have for now gone what I consider the "safe" route. We are backing up the live datasets to a B2 bucket with retention turned off. So we can prioritize what directories we pull during restoration and there are no concerns regarding versioning creating weird folder structures. I'd originally intended to include snapshots along with the live data, however, despite casting a wide net with my Cloud Sync Task, they aren't currently being captured. I probably just need to deviate from defaults and unhide the Snapshot storage.

This lets us feel safe about going live with the unit and starting the process of migrating our files from our old server but my plan going forward, assuming no one is able to give me some informed suggestions, is to later experiment with a small test dataset and bucket to see what happens with different configurations of the bucket and task, then migrate any positive discoveries into our production sync.

 

[adrianwi]

One of the problems I encountered using the Cloud Sync Tasks to Backblaze on a production server is where the data is changing during the course of a backup. I was trying to back up a Home Assistant database and that gets written to almost constantly, so errored with a mismatch when trying to sync. Backing up the local backup which only changes daily was a better solution for me.

 

[KazuyaDarklight]

Is your local backup using Rsync or Replication Tasks? Based on the documentation we went with Replication, but my reading of it leads me to understand we subsequently don't have normal files to work with in the backup, just snapshots.

 

[adrianwi]

My production server replicates (ZFS reoplication) to a backup server. I then use Cloud Sync on the backup server to Backblaze. Still testing this at the minute and as my main offsite backup is still crashplan, but the 75% offer has not expired so it's costing me $10/month. Still pretty good for over 2TB of storage, but I'm backing up far more than I need to, and if I was a little more selective Backblaze would probably be cheaper and more integrated as I wouldn't need to be running a VM for crashplan.