Re: Dom0 memory allocation?

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Thu, Mar 12, 2015 at 04:36:45PM +0100, tg%gmplib.org@localhost wrote:
> Our new machine (Xeon Ivy Bridge X6, 96 GiB ECC RAM, SSD disks) which
> runs a bunch of services split over several NetBSD PV DomU, became
> messed up last might.
> 
> The hardware runs bleading-edge NetBSD 7.0 BETA and Xen 4.5.  Perhaps an
> unwise production choice, but there you go.
> 
> There are around 20 DomU systems, mostly PVs (running NetBSD and
> GNU/Linux) but also a few HVMs (the latter running FreeBSD and OpenBSD
> for software testing purposes).
> 
> The symptom of last might's problems was that both the guests and the
> Dom0 was unresponsive.  I could not run 'top' or 'ps', and /tmp (64 GiB
> tmpfs) was "full" but in reality empty.  I managed to unmount /tmp and
> therefter got enough control to shut things down gracefully.
> 
> Now, I would like to understand what might have happened.
> 
> When I tried commands from the shell (like top and ps) I got "Killed"
> from the shell.  This smells VM shortage to me.  So perhaps this might
> be as simple as inadequate Dom0 RAM allocation allocation?

Yes, it looks like ram shortage

> 
> The system has 96 GiB RAM and 20 DomU's.  I had assigned 1 GiB for Dom0
> when this happened, and no swap partition.  I have never seen any
> recommendations on how to make an informed choice of Dom0 RAM
> allocation.  Could you help?

It really depends on what you're doing in the dom0, and how much memory
you allocated to tmpfs. Each guest probably consumes a bit of memory in
dom0; but I don't know how much. HVM do probably need more than PV because
of the qemu-dm process and associated services (especially the framebuffer
and the VNC server - which of course also depends on the framebuffer
resolution).
How much is "a few" HVMs ?

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--