v2023-09.21

[msherman64]

It's been a while since our last "release", but we've added two main features of interest:

Large Instance Launches

Large instance launches with baremetal are much improved, leveraging the "batching" feature from upstream networking-generic-switch.
https://docs.openstack.org/networking-generic-switch/latest/configuration.html#batching

To enable:

1. enable etcd by setting enable_etcd: true in defaults.yml, then run deploy --tags etcd
2. Select etcd as the coordination backend for networking-generic-switch. This should be done automatically when you reconfigure neutron with enable_etcd set, but you can also set it manually by specifying ngs_backend_url
3. Enable batching for each switch that you'd like to use this feature: Under the switch's entry in defaults.yml, set ngs_batch_requests: true. After making the changes, run deploy --tags neutron

Initial support for ironic node inspection

We're still working on generating the reference-repository data from it, but you can now use ironic inspection as a way to "healthcheck" your nodes. As of this release, you can enable the feature by editing your ansible hosts file to add ironic under the group ironic-inspector:children as seen here, then running deploy --tags ironic

For nodes already enrolled, you'll need to update the ironic "inspect interface" by running openstack baremetal node set --reset-inspect-interface for each node.

After this, you can inspect a node by the following sequence of commands:

1. openstack baremetal node manage <node>
2. openstack baremetal node inspect <node>
3. After inspection completes, run openstack baremetal node provide <node> to make it available for use again

This procedure will exercise all features needed to boot a node, including setting the boot order on the node, baremetal switch networking, neutron dhcp config, routing from the provisioning network to the ironic pxe/ipxe service, and launching the ironic agent, only excluding the use of glance images.

Not yet included:

In a forthcoming release we will use the information gathered by inspection to verify the network port configuration on nodes, as well as auto-populate a searchable inventory and blazar reservation labels.

Note:

We've moved our image hosting to the github container registry. You'll note that docker images names have changes from docker.chameleoncloud.org/kollla/... to ghcr.io/chameleoncloud/kolla/.... This was done to improve the speed and reliability of container image hosting, and to better integrate with our CI/CD process for container image releases.

Full Changelog: v2022-10.12...v2023-09.21

---

This discussion was created from the release v2023-09.21.

[msherman64]

Inspection also gives us an easy way to test if nodes are set up/connected correctly.
The script in this Gist (for now) will set nodes to be inspected, then tail the serial console.
https://gist.github.com/msherman64/1078ff38419de08b597103259031cd09