Re: Telemetry within BOSH
Damzog Jochen (CI/OSC1)
Hi Marco and Mike,
We have since long time a requirement that is similar. I have noted it down some time ago:
Subscription mechanism for events
bosh does have a notion of events, see http://bosh.io/docs/events. These events will be fired on each activity taken by the director o.a. deployments, creation and deletion of vms. These events could very well be used to trigger operations outside of bosh to either prepare or complement the action taken by the director. For example we would like to use these events to trigger creation or deletion of firewall configurations.
In order to use events to trigger external action we require to have a mechanism to subscribe for these events. There are multiple ways to achieve this. For example, the director could forward these events to a messaging system (like rabbitmq) or it could offer a mechanism to register webhooks.
For our use case of setting up FW rules it would be useful to configure a synchronous coupling between the director action and the external action to ensure FW rules are applied before a particular vms is started. This feature, however, should remain configurable if implemented because many other use cases will probably prefer to be executed asynchronously.
Mit freundlichen Grüßen / Best regards
Von: cf-bosh@... <cf-bosh@...>
Im Auftrag von Marco Voelz
Thanks for bringing this up. More insight in what the Director could definitely be helpful for a number of things.
Concerning your use-case, I hope you can help me understand a few points:
Thanks and warm regards
I have an interesting use case in front of me that I’m trying to figure out how to approach both sanely as well as sustainably. I have a use case where I want highly structured telemetry data from the BOSH director for downstream analytics. My target goal of this use case is to have a comprehensive and clear perspective into where BOSH is spending it’s time. I could see any type of telemetry data being immensely useful for operators and CFF developers as it can help give insight into where improvements can be made.
Currently I’m exploring 3 options:
Looking across the industry, telemetry is very prevalent, and it’s almost always an opt-in model, so anything I explore . Since I haven’t seen anything like this discussed in the mailing lists before, I wanted to surface my explorations to get others’ thoughts and opinions on telemetry within BOSH.