Jan Beulich wrote:

>>>> On 30.04.12 at 13:21, Per Jessen <[email protected]> wrote:
>> I have two xen hosts -
>> 
>> host1 is 32bit, 16Gb RAM, running 12.1+updates, kernel 3.1.9-1.4-xen.
> 
> Hardly - there's no (supported) 32-bit Xen on 12.1.

I guess I must have built it from source. 

>> host2 is 64bit, 32Gb RAM, also 12.1+updates, kernel 3.3.0-2-xen.
>> 
>> dom0 on host1 boots fine in 256Mb, but for host2 I need 512M.
>> It's not a problem as such, but I'm curious as to why?
> 
> How can we know or even guess, when you neither tell us anything
> about the differences of the hosts beyond their bitness, nor describe
> (or even better provide hard data) in what way it fails with less than
> 512Mb.

The hosts are virtually the same - both quad Intel Xeon, one is
4x2.8GHz, the newer one is 4x3.0GHz.  When I started host2 with
dom0_mem=256Mb, I saw lots of oom messages and it ended with a kernel
panic.

> That said, I don't really see the point in this small a host - how
> would you expect any guests to run there (which after all is what
> you'd want to use Xen for)?

I must have misunderstood something - what is the typical or recommended
size for dom0?
 
>> I've also noticed that the virtual footprints of a few processes
>> are significantly different from host1 to host2:
[snip]
>> For instance, xend on host2 appears to have a virtual footprint
>> of more than 10 times that of xend on host1.  Is this is really
>> just due to 32 vs 64bit architecture?
> 
> Presumably not directly, but given the list of processes above I don't
> see the direct relation to Xen - apart from xend, all others aren't
> Xen specific, yet show similar patterns. For xend, given that it's a
> Python script, looking at Python's footprint in general would probably
> get you much further.

Thanks, I was thinking that too (after I'd hit send). 



-- 
Per Jessen, Zürich (21.8°C)

--
To unsubscribe, e-mail: [email protected]
To contact the owner, e-mail: [email protected]

Reply via email to