Remove query-frontend without query-scheduler about mimir HOT 3 OPEN

@pracucci -- Clearly we're not going to do this in time for the GEM 2.0 launch, but I am just curious for if/when we would do this, does this mean that now the 'query-scheduler' would be a required component for a microservices deployment and would have to become a component in the monolithic deployment (meaning that the 'query-scheduler' would now need to be included when target=all?).

from mimir.

and would have to become a component in the monolithic deployment (meaning that the 'query-scheduler' would now need to be included when target=all?).

It's tricky because the query-scheduler can become a component of -target=all but it wouldn't work fine with any large number of replicas, for the same reason why it doesn't work fine the query-frontend (number of query-scheduler replicas should be <= queriers).

from mimir.

Ahhh good point. Thanks for clarifying.

from mimir.