I'm pretty new in using MAAS (and juju) for my company internal Openstack deployment; In the past I used kolla-ansible.
More details on installation and environment: Ubuntu (server) 18.04.3 LTS (GNU/Linux 5.3.0-29-generic x86_64); MAAS version 2.6.1
We have a Dell Poweredge C6420, with 4 nodes, each of them provided with 2 Mellanox 100Gb adapters; Cause of we don't have a 100Gb capable switch to interconnect the nodes, in the past:
- First of all we have physically cross-connected the nodes
- (and then) We used Linux Bridge: Each node with a br0 and the 2 Mellanox adapters connected to it.
In this way we were able to interconnect the 4 nodes, avoiding loops by enabling STP on each br0. Everything worked fine, 4 vlans ran on that network configuration (one for Openstack internals endpoint, one for storage, one for swift, one for ceph cluster and so on).
After this (golden!) period of Kolla-ansible R&D, company decided to switch to Juju+Maas.
Now we are in trouble for replicating same configuration with MAAS (both cli and web dashboard). What we want to do (out-of-the-box) in MAAS is: configure one Linux Bridge, br0, and attach both network adapters of each node to it. But Maas permit to create one br0 for one interface and this is not exactly what we want.
We didn't found a way to do this. The only one way was to, use a cloud-init base64 encoded script and use MAAS cli (as documented on maas custom setup doc page). Unfortunatelly Curtin didn't work.
But, questions are:
- How is possible to replicate the configuration described above out-of-the-box in Maas?
- Even if we workaround the problem by using Maas cli and base64 encoded cloud-init script, when we'll deploy Openstack base bundle, is there a way to preserve this (handcrafted) network configuration? Or Juju deploy we'll destroy it?
Thanks in advance