Re: [PATCH 1 of 4] Prevent low values of max_pages for domains doing sharing or paging

[Tim Deegan <tim@xen.org>]

At 06:45 -0800 on 16 Feb (1329374751), Andres Lagar-Cavilla wrote:
> Fair enough (also referring to Jan's comments). We would be building
> policy into the hypervisor.
> 
> But I've seen squeezed set criminally low max_pages value (i.e. 256).
> Granted, this is squeezed's problem, but shouldn't some sanity checking be
> wired into the hypervisor?

Some operating systems do just fine in 640K. :)  But seriously, what
lower limit would we use?  Stupidly low max_pages for some uses would be
just fine for others. 

> Why should we even allow max_pages < tot_pages?

The reasoning is:
 - the tools want a hard guarantee that a rogue balloon driver can't 
   mess up their calculations of how much free RAM there is.
 - when a VM is ballooning down we don't want to have the tools spinning
   watching actual max_pages and adjusting tot_pages down as it
   changes.

It's not a particularly nice interface, though, and I'd be happy to see
it revert to the old one (where new max had to be <= current tot).  But
that will need a fix for Xapi/squeezed to handle the ballooning-down
case some other way.

Cc-ing xen-API.  Any XCP folks got an opinion about this?  Would it be
easy to make squeezed not need this behaviour?

Cheers,

Tim.