I set my account name to Drew R earlier and it seems to have altered the course of this
thread. But anyway. This guy Strahil replied to me (not showing below) and said to try
this:
Strahil: "Can you try setting the viodiskcache custom property to write back ?
Best Regards,
Strahil"
I cloned the windows 10 vm, set one to use the 8TB WDGold disk and one to use a 6TB WDGold
2+1arb mirror gluster.
The 8TB single disk VM ran full speed: 115MBps. When it took 19 minutes prior.
The gluster VM is still going as I type this...probably still on track for 19 minutes on a
6.8GB file (~5MBps).
115 is basically maxing the 1GBps ethernet to the remote server I'm reading from for
the file copy) And this is taking place INSIDE the VM managed by oVirt. Outside the VM I
can achieve expected speeds for the hardware in any direction. I'd test something
within the 10Gbps network but it's only oVirt manages devices and they can't even
read off each other faster than 5MBps.
So that was a great example of how a setting can have effect. So with a VirtIO-SCSI +
viodiskcache=write back on a single non-gluster disk I get basically max speed. Otherwise
- goat cheese.
So how do I fix the gluster performance - through oVirt....? (that gluster copy just
finished btw - 14 minutes - so maybe a slight improvement)