So far, based on my understanding of OpenStack Python SDK, I am able to read the Hypervisor, Servers instances, however, I do not see an API to receive and handle the change notification/events for the operations that happens on the cluster e.g. A new VM is added, an existing VM is deleted etc.
There is a similar old post (circa 2016) and I am curious if there have been any changes in Notification handling? Notifications to external systems from openstack
I see a documentation, which talks about emitting notifications over a message bus that indicate different events that occur within the service.
https://docs.openstack.org/ironic/latest/admin/notifications.html
I have the following questions:
Does Openstack Python SDK support notification APIs?
How do I receive/monitor notifications for VM related changes?
How do I receive/monitor notifications for compute/hypervisor related changes?
How do I receive/monitor notifications for Virtual Switch related changes?
I see other posts such as Notifications in openstack and they recommend to use Ceilometer project, which uses a Database. Is there more light-weight solution than using a completely different service like Ceilometer?
Thanks in advance for your help in this regard.
As far as I see and I know, Openstack SDK doesn't provide such a function.
Ceilometer will also not help you. It only collects data by polling and by notifications over RPC. You would still have to poll the data from ceilometer by yourself. Beside this, ceilometer alone has the problem, that it only grow and will blow up your database, that's why you should also use gnocchi, when you use ceilometer.
At the moment I see only the 3 possible solutions for you:
Write your own tool, which runs permanently in the background and collect the data in a regular interval over OpenstackSDK and REST-API requests.
Write something, which does the same like ceilometer by reciving notifications over oslo-messaging (RPC). See the oslo_messaging_notifications-section in the configs: https://docs.openstack.org/ocata/config-reference/compute/config-options.html#id35 (neutron has also such an option) and use messagingv2 as driver like ceilometer does. But be aware here, that not every event creates a notification. The list of the ceilometer meter-data should give a good overview of which event are creating a notification and what can only be collected by polling: https://docs.openstack.org/ceilometer/pike/admin/telemetry-measurements.html. The number of notification-events is really low, so its possible, that it doesn't provides all events you want.
Use in the oslo_messaging_notifications-section in the configs log as driver to write the notification in a log-file, and write a simple program to read the log-file and process or forward the read content. Here is the same problem like in number 2, that not every event creates a notification (log-entry in this case here). This has also the problem, that the notifications and so also the event-logs, are created on the compute-nodes (as far as I know) and so you would have to watch all compute-nodes by your tool.
Based on the fact, that I don't know, how much work it would be to write a tool to collect notifications over RPC and because I don't know, if all events you want to watch really creates a notification (base on the overview here: https://docs.openstack.org/ceilometer/pike/admin/telemetry-measurements.html), I would prefer number 1.
Its the easiest way to create a tool, which runs GET-Requests over REST-API in a regular interval and forward the results to the desired destination as your own custom notifications.