I have a dataset on the source machine that I’m syncing with the target machine in a pull-style replication.
I have a retention strategy for this dataset, but I want that strategy to be applicable on the target machine and don’t want to have additional snapshots on the source machine.
For example, I want to have 7 dailies 4 weekly and 12 monthlies for this dataset only on the target machine and not on the source.
Is this achievable?
Sure, as long as you’re taking dailies, weeklies, and monthlies on the source–and not getting rid of them until after they’ve replicated off.
Let’s say you’re replicating daily. You probably don’t need to do anything special–as long as you keep one daily, one weekly, one monthly (or more) on the source, you can set the target to keep more of them than that.
In a relatively simple setup, the default synchronization snapshots will work to make sure your source and target system don’t get out of sync with each other–because if you ever lose your last common snapshot shared by both sides, your incremental replication chain is permanently broken, and you have to start again with a full.
In more complex setups–particularly ones involving more than just two partners–the default sync snapshots become more of a pain than a blessing, and you probably want to disable them entirely with --no-sync-snap. In that case, you’ll need to simply make sure that you’re replicating frequently enough (and monitoring your replication well enough) to make sure that if you have a problem replicating, you catch it and resolve it before the source destroys your last common daily/weekly/monthly.