ZFS replication is a very cool and useful feature. Have not used it in a production environment but only on my home ZFS storage server and it has got me out of problems a couple of times.
Basically as far as I am aware zfs send and zfs receive will send the named zfs snapshot across to another host running the zfs receive command and will then provide an exact copy of the original snap.
The good stuff then happens when you run a zfs send that only sends the diferences between the original snapshot and the new snap. Only the changed blocks (I think it is blocks) are sent so traffic is minimal dependant on your environment etc.
I use netcat to send the data streams between hosts as it seemed the quickest way although there are other ways such as ssh and mbuffer etc.
An example would be assuming you have a zfs dataset called pool/data and a snapshot called data@today on the sending server and that you have a zfs dataset called pool on the receiving server:
on target server first:
nc -w 120 -l -p 9090 | zfs receive pool/data
on sending server then:
zfs send pool/data@today | nc -w 20 192.168.0.2 9090
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.