"Using the management network interface as Neutron External can result in losing external connectivity" about openstack-multinode

[MrLinks75]

Excellent, thanks a lot for the valuable help !
And yeah the instructions are mostly for the host machine so it allows me to ping the external network.

Thanks again, and great project, I'll close the issue as resolved.

[electrocucaracha]

Hi @MrLinks75,

Foremost, thanks for using this project and most important, for reporting this issue. More likely, I need to improve the documentation. The idea to consume the install.sh script directly is for bare metal deployments, I have some CI coverage of that here resulting in an execution like this one, I have to recognize that the IP management discovery process needs to be improved, specially in those cases where ip route provides a non-expected output (like more than one default route, or different IP address column).

If you installed Vagrant (my preference), you can try the different setups running vagrant up from the different samples folder, like All-in-One.

In order to specify the secondary NIC in a bare metal deployment and avoid that error, you can use the OS_KOLLA_NEUTRON_EXTERNAL_INTERFACE variable, which in your case is enp23s0. So the one-liner command should be like:

 curl -fsSL https://raw.githubusercontent.com/electrocucaracha/openstack-multinode/master/install.sh | OS_KOLLA_NETWORK_INTERFACE=eno1 OS_KOLLA_NEUTRON_EXTERNAL_INTERFACE=enp23s0 bash

[MrLinks75]

Hi @electrocucaracha,
Thanks for the quick response !

I'm interested in the project for my PhD project in cybersecurity to deploy security tools such as SIEMs or UEBAs with machine-learning capabilities for monitoring NB-IoT in "production"-like environment.

I'm currently investigating multiple methods of deployment for an entire Openstack environment and your method is extremely interesting and faster than a manual deployment.

I indeed did as you recommended and used directly the vagrant file within the samples folder, especially AIO.

The deployment went without a problem.

If a need to re-deploy everything arises I will try to do it through the install.sh since I'm working on bare metal.

However, I was wondering if it was possible to apply routing rules directly through the vagrant file?

I've encountered a networking issue in the past (specifically instances not being able to ping or access external networks / internet) on manual but also automated deployments

Could commands such as :

sudo ifconfig br-ex $EXT_NET_GATEWAY netmask 255.255.255.0 up  
sudo iptables -t nat -D POSTROUTING -s $EXT_NET_CIDR -o enp23s0 -j MASQUERADE  
sudo iptables -D FORWARD -o enp23s0 -i br-ex -j ACCEPT  
sudo iptables -D FORWARD -i enp23s0 -o br-ex -j ACCEPT

Be doable out of curiosity?

Regards,

MrLinks

[electrocucaracha]

All depends, if those instructions have to run on the host machine, they need to be adapted to Vagrant like these ones, but if they have to run as part of the provisioning method, there are multiple places to place them:

• To apply in all the nodes here
• To apply only in All-in-One setup before Kolla execution here