2.1.0 Refactoring

[HadesArchitect]

Thank you. Great work. Would still like to see some additional changes before we can publish this. Let us know if you don't agree.

Instance management:
It does works better now, but I still see that potential raise conditions could occur when initializing the session.

As I wrote earlier:

• Since you're using the datasource instance management you should initialize and cache a field on the CassandraDatasource struct. To me it seems like the clusterconfig should be initialized and cached on CassandraDatasource struct and then you'll use that to create/open a session per request and close it at the end.

We have an example, see /~https://github.com/grafana/grafana-plugin-examples/blob/master/examples/datasource-basic/pkg/main.go#L20, which I think should simplify your code for serving the plugin and handle the Cassandra connections/sessions.

In the example it uses the NewDataSource for instantiating a new datasource instance - this is called each time a datasource in created/updated in Grafana and locking/concurrency guarantees that it only executes exactly once for every datasource created/updated in Grafana . This is the recommended location for instantiating clients and such, e.g. your clusterconfig.

Logging:

• For example this line that logs, rather than using logger.Debug(fmt.Sprintf("Created datasource, ID: %d\n", settings.ID)) we recommend to use structured logging, e.g. logger.Debug("Created datasource", "uid", settings.UID)
• Would recommend using lowercase keys when logging, e.g. message instead of Message, to follow Grafana's convention.
• Inconsistent use of warning/error log levels, e.g. here warning are used and here error are used. Can you please use error level everywhere when logging an error?
• Please use debug level here

Kind regards,
The Grafana Team

[futuarmo]

Fixed in #138