Keda: From brokerList to bootstrapServers in the Kafka scaler trigger

Created on 4 Feb 2020 · 4Comments · Source: kedacore/keda

Hi all,
I can see from Git history that the brokerList was brokers in the Kafka scaler time ago.
What I would like to propose is changing it again to bootstrapServers which is the most common and recognized way to specify "bootstrap" servers on the first connection from a Kafka client.
Actually, it's really a "bootstrap" servers list because it doesn't need to contain ALL the brokers in the cluster but one or more from which the client gets metadata so that can connect to right brokers later (so it's for bootstrapping...).
The "brokerList" name could drive in the wrong direction with the meaning that you need to put ALL the brokers there, which is not true. This name is still used in the Kafka upstream project for the kafka-console-producer tool but it's going to be changed in order to align with the kafka-console-consumer which uses a "--bootstrap-servers" parameter instead.
Said that, how do you approach when changing something in the CRDs?
As far as I can see, triggers metadata is just a map so there is no validation at CRD level / API server level but just inside the scaler when it gets the metadata.
In this case, do you agree if I propose to change this trigger field in bootstrapServers?

Source

ppatierno

Most helpful comment

Yes, for this reason, I was asking how you are approaching changes like this in CRDs.
I will open a PR supporting both of them marking brokerList as deprecated for the time being.
On 2.0 release we'll get rid of it.

ppatierno on 5 Feb 2020

👍3