Well, I should have tried to answer my own question earlier... but
there is still a problem.
A quick reading of Wikipedia later, I can tell the following.
First, the anticipatory scheduler seems to be the worse choice when
dealing with raid arrays.
Now, a better choice in my case could be:
- the CFQ scheduler for the array in the dom0;
- and the NOOP scheduler for the VBDs in the domUs (seems already
there by default btw)
However even with CFQ/NOOP I keep getting soft resets, only the
sata_nv SATA ports, when under load. Any thoughs ?
Thanks to all who did not flame me on this one ...;-)
GoM
--
