ZFS send/receive resume?

[soundscribe]

I'm in the process of using zfs send/receive to transfer data from server A to server B. During a 1TB transfer, some network glitch occurred and the transfer hung; the SSH session was dead. I restarted the transfer and it started from the beginning.

Does ZFS has the ability to resume from a previous interrupted transfer of a complete stream? I've found a few tantalizing posts saying "it should work", but without any detail. I also see this was a feature request from years ago, but I can't find any indication it's made it into the code.

What do you all do when using send/receive on transfers that take days or weeks?

 

[TechIsCool]

versus just using ssh you could use mbuffer and send it. That should allow you to deal with network glitches but a full down most likely not.

From all the reading I have done, I don't think it's possible to resume transfers.

 

[gea]

Delphix is working on this
see OpenZFS - YouTube

Matt Ahrens - send & receive
OpenZFS European Conference 2015, Paris

 

[apnar]

Since you can send and receive via std out/in, assuming you have enough space on each, you could send to local file then transfer each delta via any transport method you like that supports resume and finally apply it on the other side once it's complete. Little more complicated and a few more steps but should handle resuming if that's more important to you.

 

[soundscribe]

Thanks for the responses. I solved the problem by moving the two servers on to the same 1gb segment, instead of over a slow link.

Got the initial transfer done in several hours rather than multiple days.

Looks like zfs send resume is well on the way, but probably won't make it into omnios for 6 months or so.