rhoas kafka status fails when we create kafka

[wtrocki]

Create kafka still requires us to use use command to set it as enabled

[craicoverflow]

Since the create kafka command performs the API operation async, the bootstrap server host is not available yet and so this would be missing when "using" the kafka instance.

We do not want to perform a synchronous operation either because this could take a long time depending on size/speed etc.

[wtrocki]

It will be always async. We should check status of Kafka before executing any command

[craicoverflow]

Just saying that the Bootstrap server host will almost always be empty at this stage in the creation process, so we can automatically use the Kafka cluster but will only have partial information

If we have the condition to do use it unless the host field is populated, the user will then have to mentally consider both scenarios (is the Kafka cluster in use or not).

[craicoverflow]

How about this flow:

1. Create Kafka cluster
2. Set current Kafka WITHOUT ClusterHost
3. When using connect command, if the ClusterHost is empty then use the one from the live version of the Kafka instance (taken from the API)
4. If it is not empty, use the ClusterHost defined in config instead

[wtrocki]

@craicoverflow Another idea would be to rely on the status the same way that UI does it and do not do anything with object if status is pending etc.

[craicoverflow]

Yes but then that means you will never actually use the kafka cluster automatically as the status will not be complete

[wtrocki]

Ah.. ignore m previous comment. Your flow works for me. Generally we can setup this cluster as the used one... and when it is still pending keep all commands printing - cluster is still processing

Eg. kafka topics create

Currently used Kafka is in Pending state. Please try again

[craicoverflow]

Sounds good to me

[craicoverflow]

When it comes to production, to we really need to save clusterHost and clusterName, when cluster is all that should be needed?

[wtrocki]

It was a preliminary improvement to not contact API to get it. While we should be checking status all the time the fact we need to apply the URL after makes things complex so we can always request it using URL. No strong opinion. Agree that current state/implementation might be not desired.

[craicoverflow]

For validation we should be contacting the API anyway. First to check if this Kafka even exists, second, the Name/Host may become out of sync with the server version, if another user updates these fields.

[wtrocki]

@craicoverflow Yes. This makes URL in config totally not needed at all.
My bad