Closed slfritchie closed 10 years ago
Thanks! Any thoughts on either warnings, or setting the rng_seed
to now
? Maybe partitioned seq int solves this, but it was surprising to me to run multiple instances of b_b and get the same exact keys generated from each one. Ditto running a benchmark again with what I thought was a really large keyspace, and doing a bunch of updates instead of unique PUTs.
I'd be fine with a PR to change it. TBH, defaulting to the same seed is a feature in the "repeatable" test direction. Changing it to default to a random seed is fine by me, but because it'd cause a big change in behavior, separating it out and adding warnings & bells & whistles seems advisable?
TBH, defaulting to the same seed is a feature in the "repeatable" test direction.
Agree, and that's why I'm kind of conflicted about it as well. Let's get this PR merged in, and we can tackle this other issue separately. Since it should probably go in another PR anyway.
+1 to merge.
cc: @reiddraper @ksauzz @kuenishi @shino