Closed muzarski closed 3 months ago
I've took the PR into SCT integration test, and mixed is working as expected (i.e. doesn't fail)
v2:
EnumeratedDistribution::contains
to separate impl with PartialEq + Eq
trait bounds on T
. Previously, we would require T
to be PartialEq + Eq
even for sampling methods - we obviously don't need these trait bounds to implement them.v2.1: In previous comment I mentioned that we don't need PartialEq + Eq
trait bounds to implement sampling methods. I forgot to remove them, tho...
Motivation
See the discussion: https://github.com/scylladb/cql-stress/pull/86#issuecomment-2204524664.
My last PR (https://github.com/scylladb/cql-stress/pull/86) introduced a new bug -
mixed
command tries to prepare a statement for each operation (read/write/counter_read/counter_write), even though it does not create the tables.This PR fixes it so we prepare the statements only for the operations that are going to be sampled (were specified in
ratio()
parameter by the user). We obviously then assume that the corresponding write workload was ran beforehand and that it created tables used by themixed
workload.