Covering vs non-covering index

I have a document type like

    "token":"LONG STRING UP TO 10KB"

I want to run SELECT token FROM default WHERE userId='d448e08e-5f0e-4b52-aec8-e4d3a487548b'

As token is long string (up to 10KB) , is it recommended I cover index (userId,token) ? Or I must use userId for index?

It all depends on avg size and total documents. You should try and see which one performs better.