osdir.com


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Octavia][kolla-ansible][kayobe] - network configuration knowledge gathering


Hi Mark et al,

I've added this network back to the compute node but unable to launch
instances on this flat network. Both the controller and compute node [1]
have IP interfaces on this network and bridge.

After trying to launch the instance I've searched through the neutron logs
and this log shows the fail but it's unclear to me why it is failing [2] .
Nova says "binding failed" and to check neutron logs where it states
"binding:vif_type=binding_failed".

The network is created successfully via Horizon as "flat" after deploying
this configuration to the host and openstack as "physnet3" and it is mapped
to "broct-ovs" in ml2.

I have Designate installed. Previously this skewed me because Designate the
cause of another issue I had some time back. I also have just 2 compute
nodes in separate Nova Availability Zones.

Would you mind taking a look if you have time at the log and output below
to see if I a) have something wrong with the setup which is causing the
problem? or b) does it look like Designate could be causing the problem,
either? For sanity sake I make re-deploy and remove Designate and re-test.

[1] http://paste.openstack.org/show/798830/
[2] http://paste.openstack.org/show/798831/

Many thanks for any help and guidance.

Tony Pearce



On Wed, 7 Oct 2020 at 20:14, Tony Pearce <tonyppe at gmail.com> wrote:

> Thank you Mark for taking the time to reply to me and provide me with this
> information. It's been a big help. I'll check this tomorrow.
>
> Thanks again. Have a great day and stay safe.
>
> Regards,
> Tony Pearce
>
>
>
> On Wed, 7 Oct 2020 at 18:13, Mark Goddard <mark at stackhpc.com> wrote:
>
>> On Wed, 7 Oct 2020 at 10:20, Tony Pearce <tonyppe at gmail.com> wrote:
>> >
>> > Hi Mark et al, thank you for your help the other day. I'm still a bit
>> stuck with this one and I am trying to test the octavia network by
>> deploying a regular openstack instance onto it (CentOS7) which is failing.
>> In fact, my other and currently "working" external network is also failing
>> to deploy instances directly onto this neetwork also. So I am wondering if
>> there's some other step which I am missing here. Completely forgetting
>> about the octavia network, I'm curious to understand why deploying
>> instances to an external network has always failed for me. I have a network
>> like this:
>> >
>> > Real network switch VLAN 20 ----> openstack external network "br-ex" (
>> 192.168.20.0/24)----openstack router----- openstack vxlan local network (
>> 172.16.1.0/24)
>> >
>> > I can successfully deploy instances onto 172.16.1.0/24 but always fail
>> when attempting to deploy to 192.168.20.0/24.
>> >
>> > The octavia network is almost a mirror of the above except that the
>> controller also has an IP address / ip interface onto the same. But
>> forgetting about this, would you happen to have any ideas or pointers that
>> I could check that could help me with regards to why I am unable to deploy
>> an instance to 192.168.20.0/24 network? There is a DHCP agent on this
>> network. When I try and deploy an instance using Horizon, the dashboard
>> shows that the instance has an ip on this network for a brief moment, but
>> then it disappears and soon after, fails with an error that it cannot plug
>> into it. The understanding / expectation I have is that the instance will
>> run on the compute node and tunnel the network back to the network node
>> where it will be presented onto 192.168.20.0/24. Does the compute node
>> also need an ip interface within this network to work? I ask this because
>> the octavia network did indeed have this but it was too failing with the
>> same error.
>> >
>> > Any pointers appreciated so I can try and keep my hair. Thank you :)
>>
>> For instances to be attached to provider networks (VLAN or flat), you
>> need to set kolla_enable_neutron_provider_networks to true in
>> kolla.yml. The compute hosts will need to be connected to the physical
>> network in the same way as controllers, i.e. they will have an
>> interface on the networks in the external_net_names list. To apply the
>> change, you'll need to run host configure, then service deploy for
>> openvswitch and neutron.
>> Mark
>>
>> >
>> > Tony Pearce
>> >
>> >
>> >
>> > On Mon, 5 Oct 2020 at 16:20, Mark Goddard <mark at stackhpc.com> wrote:
>> >>
>> >> Following up in IRC:
>> >>
>> http://eavesdrop.openstack.org/irclogs/%23openstack-kolla/%23openstack-kolla.2020-10-05.log.html#t2020-10-05T06:44:47
>> >>
>> >> On Mon, 5 Oct 2020 at 08:50, Tony Pearce <tonyppe at gmail.com> wrote:
>> >> >
>> >> > Hi all,
>> >> >
>> >> > Openstack version is Train
>> >> > Deployed via Kayobe
>> >> >
>> >> > I am trying to deploy octavia lbaas but hitting some blockers with
>> regards to how this should be set up. I think the current issue is the lack
>> of neutron bridge for the octavia network and I cannot locate how to
>> achieve this from the documentation.
>> >> >
>> >> > I have this setup at the moment which I've added another layer 2
>> network provisioned to the controller and compute node, for running octavia
>> lbaas:
>> >> >
>> >> > [Controller node]------------octavia network-----------[Compute node]
>> >> >
>> >> > However as there's no bridge, the octavia instance cannot connect to
>> it. The exact error from the logs:
>> >> >
>> >> > 2020-10-05 14:37:34.070 6 INFO
>> neutron.plugins.ml2.drivers.openvswitch.agent.ovs_neutron_agent [-] Mapping
>> physical network physnet3 to bridge broct
>> >> > 2020-10-05 14:37:34.070 6 ERROR
>> neutron.plugins.ml2.drivers.openvswitch.agent.ovs_neutron_agent [-] Bridge
>> broct for physical network physnet3 does not
>> >> >
>> >> > Bridge "broct" does exist but it's not a neutron bridge:
>> >> >
>> >> > [root at juc-kocon1-prd kolla]# brctl show
>> >> > bridge name     bridge id               STP enabled     interfaces
>> >> > brext           8000.001a4a16019a       no              eth5
>> >> >                                                         p-brext-phy
>> >> > broct           8000.001a4a160173       no              eth6
>> >> > docker0         8000.0242f5ed2aac       no
>> >> > [root at juc-kocon1-prd kolla]#
>> >> >
>> >> >
>> >> > I've been through the docs a few times but I am unable to locate
>> this info. Most likely the information is there but I am unsure what I need
>> to look for, hence missing it.
>> >> >
>> >> > Would any of you be able to help shed light on this or point me to
>> the documentation?
>> >> >
>> >> > Thank you
>> >> >
>> >> > Tony Pearce
>> >> >
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-discuss/attachments/20201008/fe11a5f0/attachment.html>