I need to transfer a large amount of data from one unRaid server to another.

 

They’re connected via my network.  I cannot move hard drives between machines.

 

Neither has a password enabled currently and the only login for them is “root”.

 

I'm only getting around 5-8 MB/s if I use an intermediate Vista machine to drag the files between the two machines.

 

I've disabled the parity drive on Tower for now.

 

I want to transfer from Media to Tower.

 

Media’s IP is:  192.168.0.191

Tower’s IP is:  192.168.0.190

 

What command(s) would I type at (I’m assuming) Tower’s command prompt to transfer the User share “Blu.ray” from Media to Tower? (they both have a User share called “Blu.ray”)

 

I already have some data on Tower so I don't really want to do a disk copy, I'd rather copy from User share into User share...

 

I’ve found the following posts and have tried adapting them for my needs but haven’t been able to get them to work.

http://lime-technology.com/forum/index.php?topic=3275.0

http://lime-technology.com/forum/index.php?topic=3204.msg27073#msg27073

http://lime-technology.com/forum/index.php?topic=2971.msg24507#msg24507

 

If you could give me step by step commands, it would be greatly appreciated.

 

Thanks in advance.

 

There is also an old How To in the Wiki:  Transferring Files from a Network Share to unRAID

 

If it does not work, hopefully one of our Linux gurus will correct it so it *does* work!

 

I have no personal experience with any of the methods.

Thanks for that link...

 

Will this still work with version 4.4 though, as smbmount calls smbs, wouldn't I need to call cifs?

 

Would this work:

 

mkdir /work      (to create a temporary location)

 

mount -t cifs //media/Blu.ray /work  (mounts the remote directory Blu.ray on media to the temporary directory work on tower)

 

cp -r /work/* /mnt/Blu.ray    (copies all files from the temp work to the share Blu.ray on tower)

 

umount /work  (unmounts work)

 

rmdir /work  (deletes work)

 

 

Thanks for that link...

 

Will this still work with version 4.4 though, as smbmount calls smbs, wouldn't I need to call cifs?

 

Would this work:

 

mkdir /work      (to create a temporary location)

 

mount -t cifs //media/Blu.ray /work   (mounts the remote directory Blu.ray on media to the temporary directory work on tower)

 

cp -r /work/* /mnt/Blu.ray    (copies all files from the temp work to the share Blu.ray on tower)

 

umount /work   (unmounts work)

 

rmdir /work   (deletes work)

 

 

I think the share "Blu.ray" will be at /mnt/user/Blu.ray, not at /mnt/Blu.ray.

 

The copy command would therefore be:

cp -r /work/* /mnt/user/Blu.ray

 

Thanks so much RobJ and JoeL, your combination of advice did it.  I'm currently doing the transfer, by my rough calculations I'm getting around 150 GB of data / hour.

 

So here's an explanation of the commands for anyone having a similar problem that I'm having.

 

This will transfer from one unraid server's user share to another unraid server's user share.

 

mkdir /work      (to create a temporary location)

 

mount -t cifs //media/Blu.ray /work   (mounts the remote directory Blu.ray on media to the temporary directory work on tower)

 

cp -r /work/* /mnt/user/Blu.ray    (copies all files from the temp work to the share Blu.ray on tower)

 

umount /work   (unmounts work)

 

rmdir /work   (deletes work)

 

Generic version:

At the command prompt of the new server (either log in if you have a KVM (keyboard, video, mouse) attached or telnet into the server remotely.

 

mkdir /work      (to create a temporary location, work is just a made up name, if you happen to have a directory on your new server called work you can call it whatever you want, such as "temp", if you don't use "work" just replace anytime you see work with whatever you use instead )

 

mount -t cifs //old_server_name/name_of_share /work   (old_server_name is the name of the old server, by default this is "tower".  name_of_share is the name of the share you want to copy on the old server)

 

--After hitting enter here you will be prompted for the root password of the old server, I didn't have one so I just hit enter.

 

cp -r /work/* /mnt/user/name_of_new_server_share    (name_of_new_server_share is the name of the share on the new server you want to copy the files too)

 

--After hitting enter here, it's actually going to do the copy from the old server to the new server's share, you'll get a blinking line, just let this go until it returns you to the command line (it's going to be around 20 hrs for me).  Obviously make sure you leave both machines on.

 

umount /work   (unmounts work)

 

rmdir /work   (deletes work)

 

 

 

Hope this helps someone.  Thanks again for helping me.

Please feel free to update the wiki

42.6 MB/s is a good speed for gigE. Thast is what I get now from xp to tower. I think you should be able to get more though, at least 60-80MB/s.

The theoretical limit is 128MB/s, and the best I have ever gotten going from aix to aix server memory to memory transfer is 110MB/s. No jumbo frames, or etherchannel involved.

the best I have ever gotten going from aix to aix server memory to memory transfer is 110MB/s.

 

Did that include a switch or 2 in the path?

 

What protocol was being used for the test?  How was that speed determined?

This post is the most authoritative one we have on gigabit speed.  After reading these stats, I went back and did a little math that finds BubbaQ's numbers are on the low side.  His analysis of latency makes sense, though.

 

BubbaQ's post implies that gigabit is only 3x as fast as 100 mbit.  But even if 100 mbit were 100% efficient, that would only provide about 38 MB/sec.  I am able to get about 44 MB/sec to my array (Windows to cache disk, measured with "DU meter" on the Windows side), so that 3X number seems low.

 

The other stat is that gigabit is waiting 75% of the time and working only 25%.  At 25% of theoretical, we are at 32 MB/sec.  Again, too low based on my observations.

 

In other research I did a while back, I found a stat that gigabit ethernet was about 40% efficient.  At 40%, your max throughput would be just over 51 MB/sec.  That seems pretty reasonable based on my 44 MB/sec number taking into account some amount of disk latency / computing overhead.  Based on that stat, 44MB/sec is a pretty good number.

 

But in each case I'm going from one workstation to one server.  I am not capturing total bandwidth on the network, but total throughput of a single pipe that has the entire network at its disposal.  Based on this, it seems that 40-48 MB/sec is about the max one could expect from a home gigabit network, not 60-80.  I wish it were higher!

 

I do know that ethernet is contention based, which means that any node can blurt out a message at any time and could corrupt someone else's message.  That's why the protocol requires acknowledgements and includes retry logic.  Ethernet tends to be efficient for lightly loaded networks, but throughput bogs down as load increases due to incrreased collisions.  Something like token ring has the opposite characteristics.  You can only send a message when you have the "token".  To send a message you have to wait - which will slow performance for lightly loaded networks.  But since there is only one token, it is impossible to have contention.  So as the load increases, network throughput increases.

 

smino - I'm very interested in what you were doing / meansuring and wonder if that are any practical outcomes to help unRAID users better tune their networks.