ZFS replication woes

[aufalien]

I've a 9.1.1 box (PUSH) that I would like to replicate to a fresh 9.2.1.5 box (PULL).

My volume is foo and the dataset of BAR.

Both boxes have foo/BAR defined.

On PUSH, I have a snap shot of foo/BAR, no snapshots on PULL.

Under replication tasks;

Filesystem/Volume = foo/BAR

Remote ZFS filesystems name = foo/BAR

Initialize remote side for once is enabled

Remote host has an IP address which is valid and reachable.

I have chosen to not use a dedicated user.

I dropped into the command and followed troubleshooting on page 108-109 of the manual.

The ssh test works fine, no password is asked so my preshared key works;


ssh -vv -i /data/ssh/replication 10.0.10.99
Also doing the next test;
zfs send foo/BAR@auto-20140423.2245-2h | ssh -i /data/ssh/replication 10.0.10.99 zfs receive foo/BAR@auto-20140423.2245-2h

Yields this error;
Could not create directory '/root/.ssh'.
cannot receive new filesystem stream: destination 'foo/BAR' exists
must specify -F to overwrite it
warning: cannot send 'foo/BAR@auto-20140423.2245-2h': Broken pipe
I assume this is fine as the option "initialize remote side for once" is doing the -F.
However using the GUI, I cannot get this process to launch.
What am I doing wrong?

 

[aufalien]

UPDATE; I had to choose both recursive and initiate. Not super crazy on initiate and hope that I don't have to do that each and every time a rep is needed. I'm hoping this is the case since its an initial rep task and that no further initiates will be needed.

 

[aufalien]

Curious, its been bothering me. Why is it that I cannot replicate an individual dataset? I can rep the entire volume with that dataset, just not the dataset.

I would enjoy the granularity in being able to do this. Has any one been able to rep just a dataset and if so, can you share your settings with me?

 

[cyberjock]

If your snapshot covers a dataset, then the replication will be for just that dataset.

If your snapshot covers the pool, then your replication will be for the pool.

This is one of those "why would I want a dataset over a folder" moments.

 

[aufalien]

I redid my snapshot specifying foo/BAR as per my example.

However under replication, I chose both recursive and initialize options.

Its helps to have some one confirm so thanks for that.

 

[aufalien]

Curious, seeing that concurrent replication is currently not possible (https://bugs.freenas.org/issues/4523), what would you suggest in terms snapshot methodology?

Is it more reliable to snapshot an entire volume or individual datasets off that volume?