When I initiate a copy on midnight commander, from one internal location to another internal location, I peak out at about 20 MB/s. It seems like it should move along quicker than that when it's from one internal drive to another, especially with an SSD cache set up on the share in question.

 

Not sure what info will be needed, let me know if any additional logs need to be added to help troubleshoot.

 

Model: Custom

M/B: ASRock - C2750D4I

CPU: Intel® Atom™ CPU C2750 @ 2.41GHz

HVM: Enabled

IOMMU: Disabled

Cache: 448 kB, 4096 kB

Memory: 16384 MB (max. installable capacity 64 GB)

Network: eth0: 1000Mb/s - Full Duplex

eth1: not connected

Kernel: Linux 4.1.18-unRAID x86_64

OpenSSL: 1.0.1s

tower-diagnostics-20160331-1525.zip

What exactly are you doing? You talk about copying from one drive to another, but you also talk about cached user shares. Which is it? Cache is only involved when writing to user shares. If you write directly to a specific disk cache is not used.

I'm copying from one folder in a user share to another folder in the same user share (which has a cache set up on it) I never copy anything to a specific drive.

 

A move is instant but sometimes I need to copy instead of move.

Have you verified that the files actually wind up on cache?

I'll check, but even without cache I would think it would copy faster than 20 MB/s. Thats not even usb 2.0 speed 

 

I'll check to see where the files are landing and report back.

Well, that was interesting. I tried a copy of a large file from the user share to each disk directly and each one copied at 100 to 180 MB/s, including to the cache drive. When I switched to the user share to user share it cut down to 25 MB/s.

 

Thoughts?

Post diagnostics again. I suspect cache drive corruption. What filesystem is your cache?

Cache is two 128gb ssd drives, same model, BTRFS. Everything else is XFS. New diag attached.

tower-diagnostics-20160331-1925.zip

Well, that was interesting. I tried a copy of a large file from the user share to each disk directly and each one copied at 100 to 180 MB/s, including to the cache drive. When I switched to the user share to user share it cut down to 25 MB/s.

 

Thoughts?

You should never mix user shares and disks when copying or moving files unless you know exactly what to avoid. Search for User Share Copy Bug.

 

The recent activity didn't even register in the syslog, which isn't unusual if nothing is wrong.  Your cache is raid1? How full?

Yeah I was simply doing that for test purposes, generally I always copy to the user share and never directly to disks.

 

Yes, raid 1 Only using 6gb right now. - I'll run the move right now and pull the cache from the user share in question and see how a copy goes then, that way I can rule out or pin it to the cache.

 

I didn't allow the copy to finish during the test, maybe that's why it didnt show in the log.

Ok, SSD cache is out of the picture and a large copy is moving at about 12 MB/s.

 

This is when I copy from one place on a user share to another place on the same user share. From one drive to another it flys.

Is anyone else able to reproduce this type of slowness? Try copying a file that is 3gb or more from one place on a user share to another place on the same user share with mc. I only get 12 to 20 mb/s when I do a copy this way. Disk to disk is 100 to 200 mb/s.

 

-NeoenY152

It's really frustrating that it takes 2.5 hours to copy 150 gb internally. Am I really the only one this happens to? Could someone please try to reproduce this and at least let me know I'm not alone.

 

Reproduce by copying a large file from a user share to another place on the same user share. Take note of transfer speed and compare with expected transfer speed (drive to drive). I get 100 to 200 MB/s drive to drive and only 10 to 20 MB/s share to share. All internal.

Still wonder if there isn't some filesystem corruption on some disk that is part of the user shares you are having an issue with. Maybe check all disks filesystems.

What is strange though is that I can copy at expected speeds if I bypass the user share. Have you tried a copy from user share to user share to see if it happens to you?

 

It seems like to me that it will act the same on all unraid instances but I'd like to confirm that. My drive's SMART tests come back fine and parity check comes back fine with 0 errors. Is there further checking I can do on my disks?