cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marcus Sorensen <shadow...@gmail.com>
Subject Re: CentOS 6.5 | KVM | Cloudstack 4.2
Date Fri, 24 Jan 2014 22:37:18 GMT
Let's say you have eth0. You've configured your switch to share the
following vlans:

vlan default = management
vlan 200 = public
vlan 300-500 = guest

create your cloudbr0 with eth0, this has your management ip. Point all
traffic types to cloubr0 via traffic label. you should be done.
Cloudstack will bring up eth0.200 and the bridge for it, and any guest
bridges as they're assigned.

If your management network is also tagged, then create eth0.<mgt tag>
and put the bridge on that.

On Fri, Jan 24, 2014 at 3:18 PM, Maurice Lawler <maurice@daoenix.com> wrote:
> Ah, so I would revert to what I said previously; create eth0.100 etc, and
> then create cloudbr0 for all communication, correct?
>
> On 1/24/14, 5:15 PM, Marcus Sorensen wrote:
>>
>> tagged network is when you send multiple vlans to the same interface,
>> e.g. eth0.100
>>
>> On Fri, Jan 24, 2014 at 3:09 PM, Maurice Lawler <maurice@daoenix.com>
>> wrote:
>>>
>>> Tagged network, I am not sure what you mean by that; is that what
>>> Cloudstack
>>> will do once I populate everything within the UI?
>>>
>>> Along with that, making the bridges will also be done via the cloudstack
>>> code, as you mentioned.
>>>
>>>
>>>
>>> On 1/24/14, 4:21 PM, Marcus Sorensen wrote:
>>>>
>>>> Yes, assuming you have tagged networks. Just create a cloubr0 ONLY,
>>>> and use that as the traffic label for everything. cloudbr0 should be
>>>> on your 10.x network, assuming that it is the internal mgmt network,
>>>> with an ip. Then when you fill out your public network info, provide
>>>> the vlan tag. The code should look at the label, see cloubr0, find the
>>>> parent device, create an eth0.<publictag>, and a bridge for it. It
>>>> will do similar for the guest networks.
>>>>
>>>> On Fri, Jan 24, 2014 at 2:06 PM, Maurice Lawler <maurice@daoenix.com>
>>>> wrote:
>>>>>
>>>>> This may be a rather ridiculous question.....
>>>>>
>>>>> I have two subnets: 96.x public and 10.x private - What I am trying to
>>>>> accomplish on one NIC / KVM / CentOS, this can be done right?
>>>>>
>>>>>
>>>>>
>>>>> On 1/24/14, 3:41 PM, Marcus Sorensen wrote:
>>>>>>
>>>>>> You could also try these:
>>>>>>
>>>>>> This would just be an example setup to use, with management on cloubr0
>>>>>> and public on cloubr1:
>>>>>>
>>>>>>
>>>>>>
>>>>>> https://cwiki.apache.org/confluence/display/CLOUDSTACK/Creating+the+devcloud-kvm+environment+from+scratch
>>>>>>
>>>>>> See bottom of page for graphic depicting layout:
>>>>>> https://cwiki.apache.org/confluence/display/CLOUDSTACK/devcloud-kvm
>>>>>>
>>>>>> You can also review these, they spell out the exact settings through
>>>>>> the zone config for one-nic and two-nic configs, you could substitute
>>>>>> your own.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> http://marcus.mlsorensen.com/cloudstack-extras/cs-4.1-kvm-networking-one-nic.rtf
>>>>>>
>>>>>>
>>>>>>
>>>>>> http://marcus.mlsorensen.com/cloudstack-extras/cs-4.1-kvm-networking-two-nic.rtf
>>>>>>
>>>>>> On Fri, Jan 24, 2014 at 1:29 PM, Marcus Sorensen <shadowsor@gmail.com>
>>>>>> wrote:
>>>>>>>
>>>>>>> They don't technically need ips just for VM traffic, it totally
>>>>>>> depends on your setup. You need to decide where your management
>>>>>>> network is connected and add the ip there, whether it's cloubr0,
>>>>>>> cloudbr1, or some other interface.
>>>>>>>
>>>>>>> On Fri, Jan 24, 2014 at 1:26 PM, Maurice Lawler <maurice@daoenix.com>
>>>>>>> wrote:
>>>>>>>>
>>>>>>>> The document states, create cloudbr0 and cloudbr1 without
IPs, I did
>>>>>>>> as
>>>>>>>> it
>>>>>>>> told me which didn't seem right to begin with.
>>>>>>>>
>>>>>>>> DEVICE=eth0
>>>>>>>> HWADDR=00:04:xx:xx:xx:xx
>>>>>>>> ONBOOT=yes
>>>>>>>> HOTPLUG=no
>>>>>>>> BOOTPROTO=none
>>>>>>>> TYPE=Ethernet
>>>>>>>>
>>>>>>>>
>>>>>>>> DEVICE=cloudbr0
>>>>>>>> TYPE=Bridge
>>>>>>>> ONBOOT=yes
>>>>>>>> BOOTPROTO=none
>>>>>>>> IPV6INIT=no
>>>>>>>> IPV6_AUTOCONF=no
>>>>>>>> DELAY=5
>>>>>>>> STP=yes
>>>>>>>>
>>>>>>>> DEVICE=cloudbr1
>>>>>>>> TYPE=Bridge
>>>>>>>> ONBOOT=yes
>>>>>>>> BOOTPROTO=none
>>>>>>>> IPV6INIT=no
>>>>>>>> IPV6_AUTOCONF=no
>>>>>>>> DELAY=5
>>>>>>>> STP=yes
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> On 1/24/14, 3:23 PM, Marcus Sorensen wrote:
>>>>>>>>>
>>>>>>>>> so...
>>>>>>>>>
>>>>>>>>> eth0 -> cloudbr0 ? And that's the management interface?
If so,
>>>>>>>>> where
>>>>>>>>> is
>>>>>>>>> the ip for the server? I don't see any ip on cloudbr0,
that might
>>>>>>>>> be
>>>>>>>>> why you
>>>>>>>>> have no access.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> On Fri, Jan 24, 2014 at 12:38 PM, Maurice Lawler
>>>>>>>>> <maurice@daoenix.com
>>>>>>>>> <mailto:maurice@daoenix.com>> wrote:
>>>>>>>>>
>>>>>>>>>        Marcus,
>>>>>>>>>
>>>>>>>>>        So I have gone through the docs and set it up
as discussed.
>>>>>>>>> I
>>>>>>>>> am
>>>>>>>>>        now unable to gain access to the server:
>>>>>>>>>
>>>>>>>>>        The screen shot I have here:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>        That shows you cloud0 which was setup automatically,
>>>>>>>>> cloudbr0
>>>>>>>>> and
>>>>>>>>>        cloudbr1 which I setup both, of course both without
IP
>>>>>>>>> address,
>>>>>>>>> as
>>>>>>>>>        it states to do in the docs. Along with that,
I have eth0
>>>>>>>>> setup
>>>>>>>>> as
>>>>>>>>>        bridge, eth0.100 - eth0.300 setup according to
the docs. The
>>>>>>>>>        eth0.100 has the public facing IP address, however,
my
>>>>>>>>> connection
>>>>>>>>>        times out; I saw other examples where the public
IP address
>>>>>>>>> was
>>>>>>>>>        attached to cloudbr0, can you please tell me what
I am
>>>>>>>>> missing?
>>>>>>>>>
>>>>>>>>>        - Maurice
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>        On 1/24/14, 12:04 AM, Marcus Sorensen wrote:
>>>>>>>>>>
>>>>>>>>>>        I've always setup cloudbr0 (pub/mgt/guest
br) per the
>>>>>>>>>> documented
>>>>>>>>>> examples,
>>>>>>>>>>        and never cloud0 (link local bridge). You
can look at the
>>>>>>>>>> devcloud-kvm doc
>>>>>>>>>>        for an example of an all-in-one. The traffic
labels
>>>>>>>>>> reference
>>>>>>>>>> bridges, so
>>>>>>>>>>        you have to have a bridge to enter as a traffic
label in
>>>>>>>>>> the
>>>>>>>>>> first
>>>>>>>>>> place.
>>>>>>>>>>        If you don't provide traffic labels, it by
default looks
>>>>>>>>>> for
>>>>>>>>>> cloudbr0
>>>>>>>>>> for
>>>>>>>>>>        public and cloudbr1 for guest and private.
>>>>>>>>>>
>>>>>>>>>>        Looking through the code, it looks as though
if you stick
>>>>>>>>>> with
>>>>>>>>>> an
>>>>>>>>>>        'untagged' public network (enter no vlan id
in your public
>>>>>>>>>> range),
>>>>>>>>>> then
>>>>>>>>>>        you're required to create the bridge yourself,
matcing the
>>>>>>>>>> traffic
>>>>>>>>>> label
>>>>>>>>>>        you enter. If you enter a vlan id, then it
will create the
>>>>>>>>>> public
>>>>>>>>>> bridge
>>>>>>>>>>        for you, but you still have to identify where
you want the
>>>>>>>>>> bridge
>>>>>>>>>> to
>>>>>>>>>> be
>>>>>>>>>>        created via traffic label. e.g. say you have
only cloudbr0,
>>>>>>>>>> which
>>>>>>>>>> is
>>>>>>>>>> your
>>>>>>>>>>        mgmt bridge, and you want vlan 460 on that
same eth device
>>>>>>>>>> to
>>>>>>>>>> be
>>>>>>>>>> public
>>>>>>>>>>        traffic. You'd enter 460 as the vlan id when
entering the
>>>>>>>>>> public
>>>>>>>>>> traffic
>>>>>>>>>>        range, and set the traffic label to 'cloudbr0',
to identify
>>>>>>>>>> where
>>>>>>>>>> the
>>>>>>>>>> vlan
>>>>>>>>>>        460 bridge should be created. it then looks
up the physical
>>>>>>>>>> interface
>>>>>>>>>> that
>>>>>>>>>>        cloudbr0 is bridged to (eth0), creates a tagged
interface
>>>>>>>>>> (eth0.460),
>>>>>>>>>> and a
>>>>>>>>>>        bridge (breth0-460).
>>>>>>>>>>
>>>>>>>>>>        For private traffic (mgmt), it expects you
to have already
>>>>>>>>>> created
>>>>>>>>>> the
>>>>>>>>>>        bridge. I believe this is most likely because
they expect
>>>>>>>>>> this
>>>>>>>>>> to
>>>>>>>>>> be
>>>>>>>>>> how
>>>>>>>>>>        you're reaching the server in the first place
(via ssh on
>>>>>>>>>> mgmt
>>>>>>>>>> net).
>>>>>>>>>> Guest
>>>>>>>>>>        networks are always dynamically created.
>>>>>>>>>>        On Jan 23, 2014 9:11 PM, "Maurice
>>>>>>>>>> Lawler"<maurice@daoenix.com>
>>>>>>>>>> <mailto:maurice@daoenix.com>  wrote:
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>>        Hello,
>>>>>>>>>>>
>>>>>>>>>>>        I am setting up KVM / Cloudstack all under
one server. I
>>>>>>>>>>> have
>>>>>>>>>>> done
>>>>>>>>>>> this
>>>>>>>>>>>        countless of other times, however, this
time on a new
>>>>>>>>>>> server
>>>>>>>>>>> I
>>>>>>>>>>> have
>>>>>>>>>>> noticed
>>>>>>>>>>>        it did not provision cloudbr0 / cloud0
as it has done in
>>>>>>>>>>> the
>>>>>>>>>>> past.
>>>>>>>>>>>
>>>>>>>>>>>        I saw a few tutorials where it says to
setup VLANS
>>>>>>>>>>> ifcfg-eth0.100-300
>>>>>>>>>>>        which I understand. However, right now
I am not sure if
>>>>>>>>>>> this
>>>>>>>>>>> is
>>>>>>>>>>> the
>>>>>>>>>>> normal
>>>>>>>>>>>        for 4.2 to not have those two previously
mentioned
>>>>>>>>>>> interfaces
>>>>>>>>>>> already setup
>>>>>>>>>>>        when you issue the command setup-management
/
>>>>>>>>>>> setup-databases
>>>>>>>>>>> as
>>>>>>>>>>> it
>>>>>>>>>>> has
>>>>>>>>>>>        done before.
>>>>>>>>>>>
>>>>>>>>>>>        Can someone explain this to me?
>>>>>>>>>>>
>>>>>>>>>>>        - Maurice
>>>>>>>>>>>
>

Mime
View raw message