Simultaneous disk access kills the array

[tucansam]

My config right now is simple: 8 disks on an LSI controller, and one disk (for dockers, all apps, etc) on the motherboard's controller.  All channels across the board are SATA II.  This is an i5-2500 with 8 GB RAM, if it matters.

 

I am copying files to the array (no cache disk) from another machine.  Simultaneously, I am listening to a song from the server using VLC on my desktop.  My copy is chugging along at 40MB/s.

 

I just clicked to update two docker images.  The file copy slowed to 3MB/s, then 400kbps, and the song stopped playing.  The update for the docker images, which usually takes seconds, is taking minutes.  At the same time, the web GUI, which is usually slow to respond anyhow (another issue I am curious about since my other server doesn't do it) has essentially stopped.  Additionally, any directory browning I do from my desktop using Windows Explorer also seems to interrupt or at least impede any other i/o operations.

 

Similar operations yield the same result.  A simple "ls -al" on the docker's disk (not part of the array, not any part of the file copy process) took about 30 seconds to return a result during a "chown -R" operation on a large directory on the array in another ssh session.

 

It seems that any sort of simultaneous i/o on the disks is bottlenecking somewhere.  I'm running 6 Seagate Archives and two WD Reds on the array, if that makes a difference.

 

Any ideas as to what's going on?  /var/log/syslog shows various docker operations, no errors, and nothing related to i/o.  htop shows a normal sysload for what I'm doing (I've been copying files for days and the sysload has remained pretty constant) and the CPU is happily bouncing between 2% and 50% on all cores.

 

Thanks.