Re: Instability with NetBSD Dom0

[Robert Elz <kre%munnari.OZ.AU@localhost>]

    Date:        Sun, 01 Mar 2015 00:46:10 +0100
    From:        tg%gmplib.org@localhost (=?utf-8?Q?Torbj=C3=B6rn?= Granlund)
    Message-ID:  <86385plhvh.fsf%shell.gmplib.org@localhost>

  | (3) Else, how can I place many DomUs on the disk such that the Dom0 is
  |     not involved?

I think the Dom0 will always be "involved" - just in a different way, but
I use GPT and wedges.    Each DomU gets a wedge (from the Dom0) to itself,
and puts a GPT and its own wedges inside.   Everything in the DomU runs on
raidframe (raid1's) - I also planned on running raidframe in important DomU's
(and yes, there are 4 different drives) but never got around to making that
happen (yet anyway - maybe when I upgrade to NetBSD 7 after it is released).

  |     Carving up the disk with with a DOS MBR into 7
  |     partitions, and then place a NetBSD label on each?

I think that kind of setup is supposed to work on FreeBSD, but doesn't
using NetBSD - there's only ever one (active) BSD disklabel on a disk
(visible to one kernel - client kernels can have their own obviously).

  | (4) Hand on heart, do you think I can make this system stable?  How?

I have no stability problems (aside from power outages), my DomUs tend to be
about as big as the one you're trying to make - only real difference is that
I'm using "domestic" grade equipment and normal discs (not SSDs) so it
probably isn't as fast as what you're building (which might matter.)

I'm also running older NetBSD than you - a 6.99.something (kernels), and
same userland for the DoM0 and 6.0 release userland for DomU, and Xen 4.3
I think (whatever was the most up to date stuff a year or so ago).  Aside
from the odd application like bind, there's been no need to upgrade anything.
I used to have occasional xl hangs if I was a bit careless when playing
with test DomU's (xl on the Dom0 would hang (tstile I think), so no new
DomU's until after a reboot - but everything else kept on working fine - the
Dom0 and running DomU's and interaction between them) but I am more cautious
now about waiting for operations to complete before going onto the next,
and haven't seen that issue in many months.

kre