Understanding usage/uptake through logging / authentication

[koendirckx]

We want to gauge uptake of the platform internally.

Ideally, would like to understand the following:

• cube/table
• columns being used
• authenticated user - failing that some session indicator?

As Marcin indicated on the slack channel, turnilo can spawn hundreds of druid topn queries for a single view, which would make parsing existing logs not as useful.

Would something like proxy authentication work for authenticating users?

[adrianmroz]

As Marcin indicated on the slack channel, turnilo can spawn hundreds of druid topn queries for a single view, which would make parsing existing logs not as useful.

But that happens on backend. We could introduce log levels and log not only druid queries but also plywood queries from frontend. We could recover a lot of information from them - at least data source and basic measures.

[mkuthan]

I would start with Apache access log like format: https://httpd.apache.org/docs/2.4/logs.html#accesslog. But how to present request from the client? Hashed Turnilo URL would be impractical to parse during analysis. Complete view definition is perhaps too detailed, we need something in between :)

[adrianmroz]

We don’t have view definition on backend. I think we should decode datasource, and filters/splits/basic measures from query.