Large Initial Data transfer help

[frostfractal]

New user to Unraid as I'm transitioning from a dead synology unit. Current unraid setup 4x12TB 7200RPM using 1 as parity in my array. I have my backup data from my synology temporarily on a 24bay IBM storage chassis. I was able to mount the storage chassis using iscsi on a second server then use smb to the unraid share. This worked but provided somewhat lackluster performance averaging around 40MB/s. (1Gb network with 10k RPM hard drives to 7200RPM Unraid Drive) I have a cache drive ready for unraid but am not using it until I complete moving my large data moves.

I'm getting ready to move about 15TB of data now and would like to know if I would be able to get better transfer rates by connecting to my storage chassis directly from my unraid host and performing a data copy to my unraid share. I have tested and am able to use iscsi initiator to establish a connection and mount my storage chassis as a read only disk device.

In MC I can then navigate to "/mnt/disks/v3700/data" this is the folder I want to move to my unraid array. This folder would have to fill and spill over to a second drive in my array as my share is using high water allocation. Would this be classified as a disk to share that I read is not advised? What is my best path to efficiently move this 15TB folder?

[trurl]

Leaving parity out until done with initial data load would give best performance. 

 

Caching would just got in the way since it wouldn't be large enough to take it all. 

 

https://docs.unraid.net/unraid-os/manual/storage-management/#array-write-modes

[frostfractal]

I have paused the pairity job until I have the initial data load complete and am still around 50MB from the VM on the server to array. Is that the best performance I will have for now?

[itimpi]

11 minutes ago, frostfractal said:

I have paused the pairity job until I have the initial data load complete and am still around 50MB from the VM on the server to array. Is that the best performance I will have for now?

Not sure what you mean by this as there is no ‘parity job’ to pause as parity in Unraid is real time).   If you have parity assigned then it is used.   If it is not assigned then you get faster transfer speeds.

 

if you mean that you are currently syncing (or checking) parity this will severely degrade any writing of new files until it is finished.

[Kilrah]

Enable turbo writes.

[frostfractal]

@itimpi yes that is what I mean. I have parity on the array but have paused the parity check until initial transfer is complete.

@Kilrah I have already enabled "reconstruct write" in the disk settings page. if this is what you are referring to.

 

I guess to summarize my initial question is getting 50MBps the most that I will get or will I get better speeds not going through a host machine vs using iScsi to transfer directly in unread in the shell with mc.

[itimpi]

The way to get better speeds for initial load is to not have any parity drive assigned at that point.   However that does mean the data is not protected against drive failures until you DO get around to having a parity drive and building the initial parity.

 

50MBps seem slow if you are not doing any parity actions and also have Turbo Write enabled - I would expect more like 80-100MBps unless there is something else going on that is contending for the drives.

[JorgeB]

You can also post diagnostics saved during the transfer in case there's something visible there.

[frostfractal]

@JorgeB Which of these log files would be most useful?

[JorgeB]

Post the complete zip.

[frostfractal]

Here is the .zip file posted.

smnas-diagnostics-20240115-1206.zip

[JorgeB]

 [name] => parity
 [status] => DISK_INVALID

 

You need to finish the parity sync first, or turbo write won't work correctly.

[frostfractal]

Ok thank you. I've got 5TB left of this large initial load (~24h based on current speeds). Guess I should have waited until after I had parity before I started. I'll do a test after this transfer is done and I get a parity sync and see if my speeds increase.