yes, having Drill process being the owner of the metadata cache makes sense since the goal of the cache is to speed up planning time. However Drill should not leak any information the user doesn't have access to as a result of this. Specifically we need to ensure the metadata queries/queries when impersonation enabled return consistent results whether the cache is enabled or not.
-Neeraja On Mon, Sep 21, 2015 at 10:05 PM, Jacques Nadeau <[email protected]> wrote: > I don't know if this is how it was specified but this makes sense to me. > The metadata cache is Drill's information, not the user's. > > -- > Jacques Nadeau > CTO and Co-Founder, Dremio > > On Mon, Sep 21, 2015 at 9:52 PM, rahul challapalli < > [email protected]> wrote: > > > This is with impersonation enabled. > > > > On Mon, Sep 21, 2015 at 7:39 PM, mehant baid <[email protected]> > > wrote: > > > > > Is impersonation enabled when you perform the refresh? > > > > > > On Monday, September 21, 2015, rahul challapalli < > > > [email protected]> > > > wrote: > > > > > > > Hi, > > > > > > > > With the newly checked-in refresh metadata cache feature, I see that > > the > > > > cache file is always created as the user who started the drillbit > > process > > > > and has nothing to do with the user who has issued the "refresh table > > > > metadata" command. Can someone from the dev verify this? > > > > > > > > - Rahul > > > > > > > > > >
