FWIW:

Its actually because of memory over-committment using ballooning virtio
device.  In theory the VM should get back the missing memory when
needed, assuming its available elsewhere on the host!!!

Daniel.



On 22/01/14 19:20, Craig Sanders wrote:
> On Wed, Jan 22, 2014 at 04:21:42PM +1100, Daniel Jitnah wrote:
>> I have tried that (free --si) , it reports 778 (or something like
>> that_.  Btw its Debian 7.
>>
>> The value 761 also seems like a weird number?
>>
>> Has anyone had any issue with Debian reporting wrong memory?
> 
> some vaguely relevant thoughts in no particular order:
> 
> 
> the kernel always uses some ram for itself, and any data in tmpfs
> filesystems will use ram too.  
> 
> 220MB or 240MB out of 1GB sounds like a lot, though....is the initrd
> still mounted? or maybe some ram is being reserved for a video card or
> other device (unlikely, most VMs only use around 10M or so for video
> emulation)?  
> 
> IMO, the most likely culprit is your VM hypervisor - kvm or xen? or
> maybe your kernel version - do you get different values for 'free' with
> different kernels?
> 
> try grepping for "Memory" in /var/log/dmesg, e.g. on my 16GB system I
> get:
> 
> # uname -a
> Linux ganesh 3.10-2-amd64 #1 SMP Debian 3.10.7-1 (2013-08-17) x86_64 GNU/Linux
> 
> # grep Memory /var/log/dmesg
> [    0.000000] Memory: 16346352k/17563648k available (3642k kernel code, 
> 826012k absent, 391284k reserved, 3127k data, 920k init)
> 
> # free -m
>              total       used       free     shared    buffers     cached
> Mem:         16047      15695        351        146          0        829
> -/+ buffers/cache:      14866       1181
> Swap:         8191       3681       4510
> 
> 
> (note: "absent" memory is irrelevant and can be ignored, it just
> indicates gaps in the motherboard's memory map...it's not actually
> missing memory - for details, see
> http://serverfault.com/questions/220626/debian-squeeze-and-available-memory-1gb-absent
>  ).
> 
> some other words to grep for are (case-insensitive) "mem" and "ram".
> 
> 
> also, has the VM been allocated 1000M or 1024M? it's not that great a
> difference, but it all adds up.
> 
> 
> craig
> 
_______________________________________________
luv-main mailing list
[email protected]
http://lists.luv.asn.au/listinfo/luv-main

Reply via email to