| 1 |
On 1/28/2013 17:52, Michael Mol wrote: |
| 2 |
> On Mon, Jan 28, 2013 at 5:35 PM, Randy Barlow |
| 3 |
> <randy@×××××××××××××××××.com> wrote: |
| 4 |
>> On 01/20/2013 12:37 AM, William Kenworthy wrote: |
| 5 |
>>> So what is usually recommended and works for this scenario? |
| 6 |
>> |
| 7 |
>> I personally use a bridged interface that allows my VMs to be on the |
| 8 |
>> "physical" network. That works out pretty well. In my use case, it's |
| 9 |
>> the same subnet as the host, but it should be possible to use VLANs to |
| 10 |
>> accomplish having them on a separate subnet. |
| 11 |
I've got a Gentoo-based libvirt/qemu-kvm host running with several VMs, |
| 12 |
also using bridged TAP adapters. It works really well for servers/other |
| 13 |
"always on" systems that run in the background. virt-manager can handle |
| 14 |
everything for you, you just have to know the name of the bridge to |
| 15 |
which you want to the VM to join. |
| 16 |
> |
| 17 |
> There's no requirement that they be on separate layer 2 segments if |
| 18 |
> you want them to be on separate layer 3 subnets. |
| 19 |
> |
| 20 |
> Either statically configure the IPs, or: |
| 21 |
> |
| 22 |
> For IPv4: Have DHCP grant IPs from different pools based on source MAC |
| 23 |
> or declared hostname. |
| 24 |
> |
| 25 |
> For IPv6: Use DHCPv6 rather than SLAAC, and follow the same principles |
| 26 |
> as for DHCP-for-IPv4. |
| 27 |
> |
| 28 |
> Sure, giving them separate layer 2 segments helps encapsulation (and |
| 29 |
> may make things easier from an autoconfiguration standpoint, |
| 30 |
> depending), but it's not strictly necessary from a technology point of |
| 31 |
> view. |
| 32 |
> |
| 33 |
While that's all true, I personally think 802.1Q VLANs are *much* easier |
| 34 |
to configure than DHCP and especially DHCPv6. Definitely sysadmin's |
| 35 |
prerogative, though. |
| 36 |
> -- |
| 37 |
> :wq |
| 38 |
> |
| 39 |
:x |
| 40 |
|
| 41 |
-- |
| 42 |
♫Dustin |
Archives