Since logstash is a stager, it will accept data, augment and prep it and ship it to elasticsearch, elastic shouldnt have to format or handle anything additional as it just verifies and adds documents. I would think that logstash SHOULD know all the datatypes of elasatic and be able to pass metadata about types accordingly.
So, how does dynamically creating indices work? I have daily indices being created, but with your logic, i would have to have a cron job to every day on a worker machine, roll over all new indices and create this template? Seems like a lot for something that should be handled?
FOR EXAMPLE: It looks when you define a output as elasticsearch, there is a manage_template option you can turn on, as well as a template and template_name property you can assign. This gives me some insight as a consumer that you can define a TEMPLATE which will be passed into Elasaticsearch.
That way you dont have elastic definition incorrect datatypes, strings where they should be ints or floats etc. You can, during the logstash initial push of the creating an index feed into elastic this metadata.
I havent really looked into it personally and are trying to find better documentation on it, but it seems there is in fact a way, though im unsure of the format or where the default file should live etc.