Local replication speed issue.

[jamiejunk]

I have a server with two jbods. I want to replicate from one to the other as a local backup.
When I manually run a zfs send and receive I get good speeds.

First i run a snapshot:
zfs snapshot tank2/data@7-14-15

Then I manually copy it over:
zfs send -v tank2/data@7-10-15 | zfs receive -F tank1/groups

When looking at zpool iostat 1 I see read and write speeds over 500 meg a second.

I set up local replication through the GUI.
Replication Stream Compression: lz4
Encryption Cipher: Disabled
Remote hostname: Set to the IP address of the machine. (not 127.0.0.1)

For some reason when I this runs I'm not seeing speeds above 150 megs a second.

Anyone have any ideas on this?

 

[depasseg]

While it's replicating, take a look at the system resources. You might be maxing out the CPU.

Try disabling compression.

 

[Robert Trevellyan]

Remote hostname: Set to the IP address of the machine. (not 127.0.0.1)

Why? Did you try with localhost instead?

 

[jamiejunk]

No I haven't. I wasn't sure that would work.

 

[Robert Trevellyan]

No I haven't. I wasn't sure that would work.

Using localhost is the correct way to replicate locally. I expect it will solve your throughput issue.

 

[cyberjock]

There are unconfirmed rumors that there may be a bug in the current version(s) of FreeNAS with regards to replicating over the network. You may be replicating entirely locally, but you are still using SSH and networking to pass the zfs send and zfs recv to each other, so the bug may be affecting you.

In the big picture I wouldn't worry about it and just let it run. Unless you are expecting all future snapshots to be massively huge this shouldn't be an issue after the initial replication is finished. Besides, since the bug isn't understood yet afaik we can't really tell you how to fix it anyway.