However, fillrandom will cause the inserted data to not fully cover the specified key range, which will result in read tests not querying the corresponding key, affecting the correctness of the experiment.
Recommend replacing fillrandom with fillseq.
Or using filluniquerandom directly to ensure the integrity of the inserted data.
Having some missing keys is also a valid test. The random fill is great for preconditioning, so we don't want to use fillseq. We could consider moving to filluniquerandom in the future through.
for
long_performance
, current testing processHowever,
fillrandom
will cause the inserted data to not fully cover the specified key range, which will result inread
tests not querying the corresponding key, affecting the correctness of the experiment.Recommend replacing
fillrandom
withfillseq
. Or usingfilluniquerandom
directly to ensure the integrity of the inserted data.